[Ffmpeg-devel-irc] ffmpeg-devel.log.20191113

[burek]

[00:40:13 CET] <cone-244> ffmpeg 03Paul B Mahol 07master:08c46e40fb68: avfilter/vF_colorbalance: rewrite, fixes filtering
[02:22:24 CET] <cone-244> ffmpeg 03James Almer 07master:245cbab556b8: avcodec/cbs_av1: support one byte long OBUs when the size is not set in the bitstream
[02:22:25 CET] <cone-244> ffmpeg 03James Almer 07master:72ec3358f44c: avcodec: add an AV1 frame merge bitstream filter
[02:22:26 CET] <cone-244> ffmpeg 03James Almer 07master:0b8742521fa2: avcodec/av1_parser: export bitstream timing info when available
[02:22:27 CET] <cone-244> ffmpeg 03James Almer 07master:97d9cff251ee: avcodec/libdav1d: export bitstream timing info when available
[02:22:28 CET] <cone-244> ffmpeg 03James Almer 07master:cc61466058ef: avformat: add an AV1 Annex B demuxer
[02:22:44 CET] <jamrial> BBB: ^
[02:30:09 CET] <BBB> ty!
[02:38:22 CET] <mkver> jamrial: If you (like me) are concerned about avoiding allocations, you might take a look at this patch: https://patchwork.ffmpeg.org/patch/14458/ for low-hanging fruit.
[04:54:53 CET] <mkver> michaelni: Backporting e78b0f83748f92ea9e93b21c36082e0dd04d7cb1 broke building with hardcoded-tables (see http://fate.ffmpeg.org/report.cgi?slot=x86_64-openbsd5.6-gcc4.2-conf2-n4.0&time=20191112223445). c8232e50074f6f9f9b0674d0a5433f49d73a4e50 needs to be backported, too.
[08:05:39 CET] <kurosu> wbs: I have this thing here that says ffhevc could be 20-40% faster but very unlikely ffhevc will actually be
[08:07:10 CET] <JEEB> would first of all require seeing any useful stuff from openhevc, then possible additional work
[08:07:32 CET] <JEEB> and all that requires time & effort
[08:07:51 CET] <kurosu> Yeah my point
[08:08:00 CET] <BBB> 40% o_O wow
[08:08:03 CET] <BBB> cool
[08:08:30 CET] <kurosu> I'd say 5-15% single threac
[08:08:36 CET] <BBB> ah ok, I was about to ask
[08:08:48 CET] <kurosu> Not accounting for dav1d levels of asm
[08:09:35 CET] <kurosu> And wpp+frame is another 20% on 4+ cores
[08:09:52 CET] <kurosu> If the bitstream allows, much like with tiles
[08:10:38 CET] <kurosu> Openhevc also had tile+frame, but I think it could be better
[08:11:05 CET] <BBB> dav1d's also has some left when it comes to threading, I explained that in my tlk at vdd last weekend
[08:11:20 CET] <BBB> always nice to see we all try to squeeze moar moar moar out of it
[08:11:40 CET] <kurosu> Was probably getting some rest from jet lag at that time
[08:12:12 CET] <kurosu> Was it about moving inloop filters in their thread?
[08:12:20 CET] <kurosu> *to
[08:13:55 CET] <BBB> yes
[08:14:05 CET] <BBB> so that the lag is not per sbrow, but within sbrow
[08:14:13 CET] <BBB> I think youhad a patch to do that to ffhevc also
[08:14:18 CET] <BBB> progress x,y instead of justy
[08:15:30 CET] <kurosu> I posted a proof of concept circa 2014, yes, but it was at best clumsy
[08:15:52 CET] <BBB> that's ok
[08:15:56 CET] <JEEB> :)
[08:16:10 CET] <JEEB> PoCs often are, as they're basically what they say on the tin
[08:16:12 CET] <kurosu> And there are better threading models.thag make it more complex to handle all dasex
[08:16:18 CET] <kurosu> *cases
[08:16:28 CET] <kurosu> Sorry for typos, obviously on phone
[08:17:11 CET] <kurosu> Anyway, maybe vdd 2020, who knows (this is not a taunt)
[08:17:39 CET] <BBB> I hear vdd 2020 may again be in a cool, exotic location
[08:17:46 CET] <BBB> and yes, please do, that'd be cool
[08:18:10 CET] <BBB> obviously hevc sucks :-p, but it'd be nice to see a cool hevc decoder and see if we can learn from it to make dav1d even better etc.
[08:18:27 CET] <kurosu> I don't think so
[08:18:50 CET] <kurosu> There's maybe 1 or 2% that could be brought over
[08:19:15 CET] <kurosu> But also dav1d could give ffhevc 2% too
[08:19:22 CET] <BBB> I mean concept, not code
[08:19:29 CET] <kurosu> In particular your array setting unrolling
[08:19:40 CET] <kurosu> Yeah concept
[08:20:28 CET] <BBB> :) that code is weird
[08:20:30 CET] <BBB> but it does help
[08:20:50 CET] <kurosu> I have a poor man's version in ffhevc for some hotspots
[08:21:48 CET] <BBB> but the threading also
[08:22:00 CET] <kurosu> Iirc you did move some intra info in the mvfield because it is often accessed together and it is more access friendly n general, our 
[08:22:05 CET] <kurosu> *IIRc
[08:22:09 CET] <kurosu> Oh
[08:22:24 CET] <kurosu> I never continued with that x,y progress
[08:22:46 CET] <BBB> but the concept applies to both ;)
[08:23:01 CET] <kurosu> Too hard to marry with wpp/tile+frame 
[08:25:06 CET] <BBB> yes, tile makes it very, very difficult if the postfilters cross tile boundaries
[08:25:21 CET] <kurosu> I mean on a few nights where rest was sufficient
[08:25:50 CET] <kurosu> Both behaviours are allowed in hevc (across or not)
[08:25:55 CET] <BBB> right
[08:26:16 CET] <BBB> what do people use most often?
[08:26:37 CET] <BBB> I guess in broadcast, not-across-tile is probably popular b/c it helps speed?
[08:27:35 CET] <kurosu> Depends
[08:27:48 CET] <kurosu> It can be ugly in some cases not to do it
[08:28:20 CET] <kurosu> But then you need to communicate across processors (think hw here)
[08:29:06 CET] <kurosu> BBL
[08:59:47 CET] <nullptr|> {quit
[09:33:55 CET] <JEEB> marton seems to have done some good work with mpegtsenc
[09:40:52 CET] <cone-661> ffmpeg 03Steven Liu 07master:4e0860e36305: avformat/hls: continue to play enabled playlist when have failed playlist
[10:36:27 CET] <cone-661> ffmpeg 03Limin Wang 07master:09afddf04e6f: avformat/hlsenc: free the original malloc pointer to avoid error when system function used in the following patch
[13:09:45 CET] <cone-661> ffmpeg 03Paul B Mahol 07master:d19fdc83b35c: avfilter/vf_colorbalance: add option to preserve lightness
[13:09:46 CET] <cone-661> ffmpeg 03Paul B Mahol 07master:bffd0f7b69b8: avfilter/vf_colorbalance: switch to floats
[13:09:47 CET] <cone-661> ffmpeg 03Paul B Mahol 07master:45f03cdd20c3: avfilter/vf_colorbalance: add support for commands
[15:54:01 CET] <kurosu> not many people in the -meeting room, so here goes
[15:54:23 CET] <kurosu> even if obvious, it would be good to have meetings notes from last Sunday
[15:54:30 CET] <kurosu> in particular results
[15:54:33 CET] <kurosu> sent to the ML
[15:54:53 CET] <JEEB> aye
[15:54:54 CET] <kurosu> (I did this once, but it was not really used/useful, so...)
[15:56:02 CET] <kurosu> also I was curious (maybe it's better that way) of the complete results, including not elected votes
[15:57:30 CET] <kurosu> Regarding the the social commitee, my (very personal?) feeling was that, besides people knowing the devs/community, (somewhat) outsiders could be cool
[15:58:42 CET] <kurosu> outsiders as people using the software/libraries but actually more involved elsewhere, and that can say "look, for an outsider, you look like <unfavourable thing>"
[15:58:56 CET] <kurosu> but maybe that's just me misunderstanding its role
[15:59:28 CET] <thardin> therre's a meeting channel?
[15:59:46 CET] <JEEB> at least it was utilized for the VDD meeting
[15:59:49 CET] <kurosu> there was, maybe temporary, dunno, as it is inactive anyway
[16:00:05 CET] <JEEB> there will be another 1st of dec as far as I remember
[16:01:08 CET] <kurosu> and maybe this time, inform "stakeholders"/concerned parties some time in advance, though it was cool that it actually worked out in such short order
[16:01:31 CET] <BradleyS> mentioning it here helped me, i joined and enjoyed the discussion as an outsider
[16:01:55 CET] <JEEB> yes, I think there was supposed to be an output on the mailing list, defining the next meeting as well
[16:02:04 CET] <BradleyS> mostly interested since i've been using ffmpeg since its inception, and of course we use it in handbrake
[16:04:31 CET] <kurosu> I'm undecided on how much public access is required: openness is nice, but I fear some people may feel not at ease expressing some opinions (eg unlike me that has been classified as "fool" for sufficiently long enough)
[16:05:16 CET] <BradleyS> certainly not everything needs to be open, but i find it better to err on that side
[16:06:13 CET] <BradleyS> i think once you have certain official groups more solidified, it will make sense for them to have some private communication
[16:14:11 CET] <Lynne> that meeting was pointless and nothing was decided
[16:15:41 CET] <Lynne> everyone still thinks I'm a deranged person for ever wanting anything done regarding michael
[16:15:50 CET] <jamrial> thanks for your input, Lynne. your pessimism is appreciated
[16:16:19 CET] <jamrial> you're not deranged. you're obsessed
[16:16:21 CET] <Lynne> and he's rewarded with a technical committee position, while carl gets a social committee position
[16:17:12 CET] <Lynne> jamrial: for wanting some recognition that there was something wrong and it wasn't my fault?
[16:17:26 CET] <Lynne> for not wanting to have this swept under the bridge?
[16:17:35 CET] <thardin> can we put fabrice on the social committee?
[16:17:47 CET] <jamrial> you have harassed him for weeks in a completely on sided grudge
[16:18:09 CET] <jamrial> *one
[16:18:26 CET] <Lynne> jamrial: and I wasn't harrassed? my opinions ignored, all discussions derailed?
[16:20:08 CET] <jamrial> how is that harassement?
[16:21:31 CET] <JEEB> Lynne: note that they are not the only ones on those committees - that's the point :P
[16:22:19 CET] <JEEB> if it all goes badly then we hopefully learn from that, but please do not paint it black just yet
[16:26:41 CET] <JEEB> thardin: that should be possible but I thought he was removed enough from the whole shebang that I think that's why he didn't get any larger amount of votes.
[16:27:54 CET] <thardin> he has control of the domain I think, so should probably be involved to some extent
[16:28:14 CET] <JEEB> I deffo don't disagree
[16:28:31 CET] <kurosu> social committee to me is for calling the stop or helping to fix issues that are mostly human behaviour things
[16:28:32 CET] <JEEB> tried to just explain why he most likely didn't pop into anyone's minds at that pint
[16:28:40 CET] <JEEB> yes
[16:28:48 CET] <JEEB> technical committee then is for technical decision making etc
[16:31:16 CET] <JEEB> anyways, the stuff picked for now is just for kick-start
[16:31:18 CET] <JEEB> to have a framework
[16:32:09 CET] <kurosu> yeah, it should change with more voters from more diverse horizons
[16:32:54 CET] <kurosu> s/should/seems likely to/
[16:35:40 CET] <durandal_1707> nothing in that meeting is official, that meeting is bad mouthing of ffmpeg
[16:36:42 CET] <durandal_1707> and i was completely ignored from all this nonsense, high elitist meeting bullshit
[16:37:11 CET] <jamrial> durandal_1707: why you weren't there?
[16:38:03 CET] <durandal_1707> nobody tell me anything about it
[16:38:13 CET] <durandal_1707> for obvious reasons
[16:38:35 CET] <durandal_1707> so for me, nothing was concluded in that pesky japanjeros meeting
[16:38:43 CET] <jamrial> durandal_1707: an email was sent a week ago about it
[16:38:46 CET] <Illya> it was on the ML for a while
[16:39:12 CET] <durandal_1707> no it was not
[16:39:30 CET] <durandal_1707> nothing was transparent
[16:40:32 CET] <kurosu> I believe the reasons are neither obvious nor targetting you, but I agree there could have been less room for such complains
[16:41:15 CET] <kurosu> on the other hand, yes, interested people could request more info or be there on time (that's how I got to participate)
[16:41:38 CET] <Illya> http://ffmpeg.org/pipermail/ffmpeg-devel/2019-November/252607.html is the relevant mail
[16:41:39 CET] <kurosu> *could have requested
[16:41:56 CET] <jamrial> durandal_1707: the meeting was announced a week ago, and it was streamed for those on irc
[16:42:10 CET] <jamrial> you could have been there, or here, and participated
[16:42:49 CET] <Illya> this is only temporary as well, you can be there for the next meeting
[16:43:04 CET] <jamrial> there will be another in december 1st. if you don't participate on it, then you're just complaining for the sake of it
[16:48:17 CET] <durandal_1707> jamrial: yes, i will not be on any non-transparent meeting if it is set in ridiuculous time
[16:48:55 CET] <jamrial> it's half a month until december. you're just making up excuses
[16:49:08 CET] <durandal_1707> Illya: that is too late, and nobody replied to it
[16:49:50 CET] <JEEB> it is an announcement, why would it need replies
[16:50:06 CET] <durandal_1707> anybody could announce anything then?
[16:50:35 CET] <JEEB> in theory yes, but if you do not have enough mass that ends up nada
[16:50:56 CET] <jamrial> durandal_1707: well, you'll happy to know that everything that was decided on it is temporary. you're in time to make these apparently opaque meetings transparent with your presence in future ones
[16:51:10 CET] <durandal_1707> and stil no living proof that meeding was actually held, and recorded
[16:51:20 CET] <jamrial> now you're just trolling
[16:51:36 CET] <durandal_1707> nope, i just pointed obviouls flaw
[16:54:06 CET] <JEEB> i am not sure if the thing was recorded with that iphone that did the conference, that csn be improved in the future if you wish for such improvements
[16:54:38 CET] <JEEB> but to say that there is no living proof is a bit disingineous or whatever that word is
[16:54:52 CET] <jamrial> well, be here on december 1st, save your own logs, film your monitor if you want
[16:55:02 CET] <JEEB> yup
[16:55:44 CET] <jamrial> but please, stop trying to shit on people's efforts to improve the project
[16:55:57 CET] <kurosu> And speak then, or forever hold your peace
[16:56:44 CET] <mkver> Regarding logs: Could a log of #ffmpeg-meeting be made public somewhere (preferably alongside the existing irc-archive)?
[16:57:21 CET] <durandal_1707> yes, see, nobody mentioned #ffmpeg-meeting, up until now
[16:57:38 CET] <JEEB> uhhh
[16:58:32 CET] <jamrial> durandal_1707: it was mentioned in this channel right as the meeting started, so people could join
[16:58:33 CET] <JEEB> it was most definitely mentioned whne the meeting was being set up
[16:58:36 CET] <jamrial> you'd know if you were here for it
[16:58:48 CET] <JEEB> multiple times even
[16:59:24 CET] <jamrial> you know, like every time there was a meeting in the past several years
[17:00:09 CET] <durandal_1707> nope
[17:00:11 CET] <durandal_1707> lies
[17:00:33 CET] <JEEB> so be it woth you if that is the path you choose
[17:00:39 CET] <durandal_1707> other times meeting was announced also on ML, several days before
[17:00:53 CET] <JEEB> it was
[17:00:56 CET] <durandal_1707> so nothing on that meeting is official
[17:01:00 CET] <durandal_1707> no it was not
[17:01:10 CET] <jamrial> i liked it more when your trolling attempts were joke replies to the ml
[17:01:30 CET] <durandal_1707> you picked right time frame so just filter bunch of people out
[17:02:20 CET] <kurosu> it was 9am for you afaik
[17:02:31 CET] <jamrial> it was morning for you, and late night for US people
[17:02:38 CET] <jamrial> you know who had the shittiest time? me, at 3am
[17:02:41 CET] <jamrial> and i was there
[17:03:05 CET] <durandal_1707> and some stuff i already hear here, that is concluded on that meeting, is so devastating
[17:03:50 CET] <Illya> whats so bad?
[17:04:01 CET] <durandal_1707> all about Carl
[17:04:47 CET] <Illya> all about? could oyu be more specific?
[17:05:44 CET] <JEEB> right. so you have an issue with a voted individual. now that is why that is not asingle person. and not even three.people
[17:05:44 CET] <durandal_1707> i have not enough details, so can not be more specific, awaiting that "meeting" logs
[17:07:03 CET] <Illya> JEEB: did the camera not record or something/
[17:07:09 CET] <JEEB> it did
[17:07:14 CET] <durandal_1707> fake camera
[17:07:26 CET] <kurosu> streaming at the very least, dunno about having a record though
[17:07:41 CET] <JEEB> yes, that is what I meant
[17:10:06 CET] <kurosu> in any case, there will be another meeting to change things, by voicing opinions and convincing people
[17:48:54 CET] <BradleyS> durandal_1707: i can post my log of #ffmpeg-meeting from that date if it's not elsewhere
[17:49:27 CET] <BradleyS> as long as you promise not to accuse me of fabricating it :P
[17:50:25 CET] <durandal_1707> lol
[17:51:14 CET] <durandal_1707> i would rather wait for official statement of matters
[17:52:08 CET] <BradleyS> times are local UTC-4: https://www.dropbox.com/s/bb6fz1joy88lya2/2019-11-10.log?dl=0
[17:52:56 CET] <BradleyS> that is my complete znc log for the day
[17:53:26 CET] <mkver> BradleyS: Thanks.
[17:53:47 CET] <BradleyS> you're welcome, glad i could join
[17:54:02 CET] <durandal_1707> ban him ^
[18:08:07 CET] <cone-320> ffmpeg 03Baptiste Coudurier 07master:11a38be99cd0: avformat/mxfenc: correctly set width values for dvcprohd
[18:13:22 CET] <cone-320> ffmpeg 03Baptiste Coudurier 07master:a1403032c8cb: avcodec/dvenc: support encoding dvcprohd
[18:18:04 CET] <durandal_1707> as expected, this is all to take money out of spi
[18:18:20 CET] <durandal_1707> and TC and CC are hilarious
[18:19:45 CET] <durandal_1707> people that havent committed anything to ffmpeg in years
[18:22:13 CET] <JEEB> yea, sure. every one of those people. not committed in years
[18:23:14 CET] <JEEB> also I can't speak of anyone else but I have no idea what any money situation is with regards to FFmpeg. My reason for taking part was the dumb idea of trying to make things better.
[18:23:50 CET] <JEEB> also as noted, dec 1st should have a meeting where you can bring out all of your grievances
[18:24:07 CET] <JEEB> and even with that, the whole setup is temporary
[18:59:42 CET] <TekniQue> JEEB: I found an extremely simple workaround for my H.264 bytestream problem. Turns out there's a parameter I can pass to ffmpeg that forces it. -bsf:v h264_mp4toannexb
[19:01:06 CET] <cone-320> ffmpeg 03Derek Buitenhuis 07master:1354c39c78e5: librav1e: Don't make users explicitly set -qp -1 to use bit rate mode
[19:01:51 CET] <JEEB> TekniQue: yes that forces annex b. I did mention that as something you can also add automagically if your writer requires it :P
[19:30:20 CET] <BBB> durandal_1707: if you disagree with TC and CC, please participate in the dev meetings and air your grievances. remote participation over irc is possible, there was a video feed last time
[19:34:43 CET] <kurosu> wow you guys are staying that long in Japan? Good night btw
[19:35:11 CET] <kurosu> err, too late for that, seems like you have a plane today
[20:00:20 CET] <BBB> kurosu: indeed 8)
[20:32:23 CET] <thardin> JEEB: FFestivus
[20:55:04 CET] <durandal_1707> TC and CC are just to make stuff less democratic, and far more elitistic
[21:01:37 CET] <nevcairiel> voting for leaders is fully democratic, direct democracy where everyone votes on everything is just chaos
[21:37:04 CET] <thardin> in my experience with föreningar, whenever something has to go a vote you've already failed
[23:46:14 CET] <cone-974> ffmpeg 03Michael Niedermayer 07master:8d5f9daacd81: tests/ref/vsynth: add missing reference files, fix build
[23:46:14 CET] <cone-974> ffmpeg 03Andreas Rheinhardt 07master:a7245adee3e9: avformat/id3v2: Avoid av_strdup for key and value of dict
[00:00:00 CET] --- Thu Nov 14 2019