[Ffmpeg-devel-irc] ffmpeg-devel.log.20120711

[burek]

[00:00] <ohsix> i don't know typical values, but there's probably a best practice for webm, without a keyframe all you can send is wiggly junk pictures anyways
[00:00] <teratorn> our solution is currently acceptable with a small gop
[00:00] <teratorn> well, when a new client connects, he maybe has a handful of frames to receive before the next keyframe comes along
[00:01] <teratorn> surely it's possible to re-encode content until that time
[00:01] <gnafu> What is your current GOP?
[00:01] <teratorn> it's between 10 and 25 i think
[00:01] <gnafu> So less than a second, basically.
[00:01] <teratorn> yeah it's not huge
[00:01] <teratorn> but we want to up it maybe
[00:15] <CIA-41> ffmpeg: 03Paul B Mahol 07master * rca16f88c75 10ffmpeg/libavcodec/tiffenc.c: 
[00:15] <CIA-41> ffmpeg: tiffenc: add .init
[00:15] <CIA-41> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[00:15] <CIA-41> ffmpeg: 03Paul B Mahol 07master * r0e003d8c91 10ffmpeg/ (doc/APIchanges libavcodec/avcodec.h libavcodec/utils.c): 
[00:15] <CIA-41> ffmpeg: lavc: add av_fast_padded_mallocz
[00:15] <CIA-41> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[00:15] <CIA-41> ffmpeg: 03Paul B Mahol 07master * r9b3f9f43a7 10ffmpeg/libavcodec/tiffenc.c: 
[00:15] <CIA-41> ffmpeg: tiffenc: use av_fast_padded_malloc(z)
[00:15] <CIA-41> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[02:35] <durandal_1707> michaelni: how sensible is adding FITS as task for socis?
[02:39] <durandal_1707> it could be relatively simple by using CFITSIO: http://heasarc.gsfc.nasa.gov/docs/software/fitsio/fitsio.html
[02:59] <michaelni> durandal_1707, i dont know FITS but it sounds like a good idea if you want to mentor it
[03:01] <durandal_1707> michaelni: FITS is image/data format used by ESA/NASA
[03:02] <durandal_1707> michaelni: i cant mentor because i will probably be without internet access rest of summer
[03:03] <michaelni> adding it cant hurt ...
[03:04] <michaelni> "i will probably be without internet access rest of summer" <---- :((( 
[03:20] <Compn> durandal_1707 : are you going on a vacation to anarctica ?
[03:20] <Compn> because, and i cant speak for the whole project, but ffmpeg could totally fund internet access for developer if needed :)
[03:29] Action: Compn likes to spend other peoples money :)
[04:13] <CIA-41> ffmpeg: 03Matthieu Bouron 07master * rb8a90976da 10ffmpeg/libavformat/mxfenc.c: 
[04:13] <CIA-41> ffmpeg: lavf/mxfenc: fix dnxhd codec uls and aspect ratio
[04:13] <CIA-41> ffmpeg: Reviewed-by: Tomas Härdin <tomas.hardin at codemill.se>
[04:13] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:13] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rdd0a9b78db 10ffmpeg/ (11 files in 2 dirs): 
[04:13] <CIA-41> ffmpeg: replace all __volatile__ by volatile
[04:13] <CIA-41> ffmpeg: __volatile__ can cause problems with some compilers and volatile is a standard keyword.
[04:13] <CIA-41> ffmpeg: Found-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[04:13] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[05:11] <Compn> http://torrentfreak.com/how-big-music-threatened-startups-and-killed-innovation-120709/
[14:42] <ubitux> oh i'm stupid.
[14:42] <ubitux> i needed to use helgrind with the valgrind box
[14:42] <ubitux> it seems to raise a few errors locally
[14:42] <ubitux> that's perfect.
[14:59] <highgod> Hi,all.I want to add mpeg4-part2 hwaccel to ffmpeg.But I am a new commer.I want to reference some implemented hwaccel.whitch one is the most similar with mpeg4-part2?Thanks 
[14:59] <av500> hwaccell using what hw?
[14:59] <Compn> .            // issue658 containse wrong w/h and MS even puts a fake seq header with wrong w/h in extradata while a correct one is in te stream. maximum lameness
[14:59] <Compn> hehe
[14:59] <highgod> hardware accelerate
[14:59] <av500> yes, what hardware?
[15:00] <highgod> dxva2
[15:00] <Compn> mpeg4 is similar to wmv3
[15:00] <Zor> I think av500 is asking e.g. what kind of GPU you are targeting
[15:00] <highgod> ffmpeg has implement h.264,vc1,mpeg2,wmv3
[15:00] <Compn> Zor : dxva targets all gpu tho :P
[15:00] <Zor> Compn: dxva is not hardware
[15:01] <Compn> highgod : wmv3 is the closest to mpeg4 asp i think
[15:01] <Compn> i could be wrong...
[15:01] <highgod> just use dxva2,on AMD GPU HD6730
[15:01] <av500> anyway, making mpeg4 faster in 2012 is kinda pointless :)
[15:02] <Compn> highgod : did you contact some videolan devels ?
[15:03] <highgod> not yet,I read the vlc code,it just use ffmpeg api
[15:03] <highgod> +av500:why?hehe.I am a new commer,So I will have a lot of questions
[15:03] <highgod> heh 
[15:06] <Compn> feel free to ask them here or on ffmpeg-devel list :)
[15:06] <highgod> @Compn:why do you think so?mpeg4-part2 isn't similar to h264?
[15:06] <Compn> and post your code, so you do not make a mistake like the vaapi people did , which means no one will merge their code
[15:07] <highgod> Yes,I will.How can I post my code?
[15:08] <Compn> either send a patch to ffmpeg-devel, or create a git branch so we can review it online 
[15:08] <Compn> at a free git host like github or something
[15:08] <Compn> theres also google code website, or sourceforge for free hosting
[15:10] <highgod> I have added h264,vc1,wmv3,mpeg2 hwaccel to ffmpeg.exe,using it's API.And it has increase the speed to trans video on my computer.Maybe has some bugs,but it works
[15:10] <highgod> Greate
[15:11] <Compn> :)
[15:11] <Compn> it will be useful to a lot of people 
[15:11] <highgod> @Compn:our team want to add some OpenCL to ffmpeg to optimize the filter and pipline
[15:11] <highgod> Do you accept it?
[15:12] <Compn> i think theres a patch to paralellize filters on ffmpeg-devel right now
[15:12] <Compn> it hasnt been reviewed yet
[15:12] <Compn> but yes, patches to add that are welcome ! :)
[15:13] <ubitux> http://pastie.org/4237671  does this look correct to anyone?
[15:13] <ubitux> it seems to fix one of the race
[15:13] <ubitux> but i have no clue about what the code really does
[15:18] <highgod> @Compn:How can I contact to vlc devels?
[15:18] <ubitux> #videolan maybe
[15:31] <Compn> highgod: yes, try #videolan
[16:06] <nevcairiel> ubitux: i think the mutex lock should go above the two assignments, but the broadcast remain below
[16:06] <nevcairiel> that would at least make more sense
[16:06] <ubitux> ok
[16:09] <ubitux> thank you
[16:45] <ubitux> wow helgrind is quite slow :)
[16:45] <ubitux> still not done
[17:14] <rohankasat> my audio and video are out of sync 
[17:14] <rohankasat> can anybody help me ?
[17:23] <ubitux> i guess i'll use the fast option next run& :p
[17:40] <ubitux> oh it's done.
[18:03] <nevcairiel> looks like in the end every mt test failed :p
[18:58] <ubitux> well maybe there are some false positive
[19:06] <Daemon404> does anyone know if ffmpeg also fails fate on mips?
[19:06] <Daemon404> i dont see any mips instances
[19:07] <Daemon404> or rather... im blind
[19:08] <ubitux> i saw two of them
[19:08] <Daemon404> i think the latest push fixed it
[19:08] <Daemon404> i think...
[19:27] <teratorn> thats cool - mips is an interesting target for embedded systems
[19:27] <teratorn> are there any low-power mips systems that could, say, encoding full 1080i 60fps h264?
[19:28] <teratorn> s/encoding/encode/
[19:28] <teratorn> or 1080p rather
[19:30] <Daemon404> rofl
[19:30] <Daemon404> no
[19:30] <Daemon404> not even close.
[19:30] <teratorn> hahaha
[19:30] <Daemon404> well
[19:30] <teratorn> what about 24 fps :(
[19:30] <Daemon404> there are some high-end mips things liek 32-core beasts (i think?)
[19:30] <Daemon404> but in general, no.
[19:30] <Daemon404> >embedded
[19:30] <teratorn> well embeeded uses dedicated hardware coders
[19:31] <teratorn> so it hardly matters
[19:31] <Daemon404> afaik nothing mips based encodes at all
[19:31] <Daemon404> there;s a few settop boxes that use mips though
[21:03] <burek> ubitux, can I ask you for to do a small docs update? :)
[21:03] <burek> -for
[21:04] <burek> for the "-re" option, I've noticed (4th time so far) that some people don't quite understand what it is and where to use it, so I thought it might be a good idea to describe it a little bit more in the html docs
[21:05] <burek> anyway, something like this might maybe be added there: Since FFmpeg always tries to read the input(s) as fast as possible, to finish the work as quickly as it can, when you want to have any sort of real-time output (e.g. live streaming), then you need "-re" option to tell FFmpeg to slow down the reading of the input and to read the input(s) in real-time. This is only meaningful if your input(s) comes from a file. If your input(s) is coming from so
[21:05] <burek> me other live streaming source (http/udp) it will already be in real-time and FFmpeg cannot possibly read it faster anyway, so no need for "-re" option in that case.
[21:15] <ubitux> burek: ok, just a min
[21:24] <ubitux> burek: i've reworded (most likely badly), and submitted it to ffmpeg-devel
[21:46] <durandal_1707> michaelni: i'm going to push exr patch that adds support for zip1 and zip16
[22:11] <michaelni> durandal_1707, ok
[22:20] <durandal_1707> michaelni: does it makes difference if calculations vaules are cached instead of being recalculated all the time or compilers are smart enough to do this alone?
[22:22] <michaelni> if its equally clean then its better not to depend on compilers doing their job 
[22:23] <durandal_1707> what "equally clean" means?
[22:35] <burek> ubitux, thanks :) :beer: :)
[22:35] <ubitux> no!
[22:35] <burek> ok ok :)
[22:35] <burek> :apple-juice:
[22:35] <ubitux> thx e
[22:35] <burek> :)
[22:36] <burek> durandal_1707, to cache your own values you need in register vars? :)
[22:42] <durandal_1707> burek: no
[22:45] <CIA-41> ffmpeg: 03Paul B Mahol 07master * r5ee41d00f6 10ffmpeg/libavcodec/exr.c: 
[22:45] <CIA-41> ffmpeg: exr: remove dead assigment
[22:45] <CIA-41> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[22:51] <burek> no, what? :) I'm just trying to guess what michaelni might have meant :)
[22:56] <durandal_1707> i guess if patch was much more trivial i would get bunch of "useless" comments
[23:00] <CIA-41> ffmpeg: 03Paul B Mahol 07master * r74f3c53b57 10ffmpeg/ (configure libavcodec/exr.c): 
[23:00] <CIA-41> ffmpeg: exr: zip decompression
[23:00] <CIA-41> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[23:01] <durandal_1707> michaelni: ^
[23:05] <pasteeater> gah, ffmpeg.org is ignoring me.
[23:06] <durandal_1707> pasteeater: me too
[23:06] <pasteeater> curse its oily hide!
[23:20] <michaelni> durandal_1707, what i meant was that if the code is equally nice looking its better to factorize things and not depend on the compielr
[23:20] <michaelni> pasteeater, what do you mean by "ignoring me" ?
[23:20] <ubitux> timeout
[23:20] <ubitux> seems solved now
[23:22] <pasteeater> michaelni: yeah, it's working now.
[23:36] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * re3c2d83183 10ffmpeg/libavformat/avidec.c: 
[23:36] <CIA-41> ffmpeg: avidec: switch to av_assert
[23:36] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:36] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r5f298c6791 10ffmpeg/libavcodec/vorbisdec.c: 
[23:36] <CIA-41> ffmpeg: vorbisdec: switch to av_assert
[23:36] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:00] --- Thu Jul 12 2012