[Ffmpeg-devel-irc] ffmpeg.log.20120504

burek burek021 at gmail.com
Sat May 5 02:05:02 CEST 2012


[00:01] <pasteeater> esby: what's your ffmpeg ./configure?
[00:02] <goiken> pasteeater seems to work& thanks a bunch
[00:03] <pasteeater> goiken: good. old versions of external libraries can smeg things up
[00:05] <esby> pasteeater: ./configure --enable-libmp3lame --enable-libxvid --enable-libvorbis --enable-gpl --enable-libfaac --enable-libtheora --enable-zlib --disable-shared --enable-libx264  --enable-nonfree --enable-version3 --enable-libschroedinger --enable-avfilter --enable-libspeex --enable-libopenjpeg --enable-libgsm --enable-postproc --enable-pthreads --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-ffplay --enable-pthreads --prefix=/usr/local --enabl
[00:05] <shysoft> FFmpeg is running really slow for me. Could it be because I compiled it without optimizations?
[00:06] <pasteeater> esby: use a pastebin service for long pastes
[00:06] <shysoft> Is this an optimized build: "./configure --enable-gpl --enable-nonfree --extra-cflags='fPIC' --enable-swscale --enable-postproc --enable-pthreads --enable-shared"
[00:07] <pasteeater> shysoft: sounds like a good reason. why did you compile without optimizations? you mean you used --disable-optimizations?
[00:07] <shysoft> I did *not* use disable-optimizations. My question is, is it possible that it would run slow without having used disable-optimizations?
[00:08] <shysoft> Correction: that is would disable-optimizations without using disable-optimizations
[00:08] <esby> pasteeater: well it was one line, do you want the ouput of the configure ?
[00:08] <pasteeater> esby: it was one line that was cut off
[00:09] <esby> pasteeater: the full line is here http://commons.wikimedia.org/wiki/User:Esby/truc#gimp_2.8
[00:09] <pasteeater> shysoft: there is no reason for a general user to --disable-optimizations
[00:09] <esby> after "En cas d'erreur de compilation "
[00:10] <shysoft> What command line options could disable optimizations (other than --disable-optimizations)? I ask because it's running slow and I'm trying to diagnose why.
[00:14] <pasteeater> how do you know you're running slowly? did you update ffmpeg and now the performance is less?
[00:43] <esby> pasteeater: managed to solve it, had to recompile libvpx with the same flags (--enable-pic)
[01:20] <Freakshow> anyone have insight on why I would be getting [swscaler @ 0x101887200] Warning: data is not aligned! This can lead to a speedloss
[01:20] <Freakshow> on just dv input
[01:25] <iive> Freakshow: do you have any filters, like crop, expand?
[01:25] <Freakshow> nope
[01:25] <Freakshow> nothing of the sort...
[01:25] <Freakshow> it's quite odd
[01:25] <Freakshow> I'm not in a position to gather any relevant bits at the moment
[01:26] <Freakshow> I'll ping the channel later when I can share more data
[01:26] <Freakshow> just thought I would ask about it for an off the hip response
[01:26] <Freakshow> ;)
[01:28] <iive> Freakshow: you are using ffmpeg, don't you?
[01:40] <Freakshow> HA!
[01:40] <Freakshow> yes...
[01:53] <tg_> hey
[01:53] <tg_> anybody want to shed some advice on multithreading bottlenecks?
[10:44] <tetracon> hi anyone in atm?
[10:45] <tetracon> I'm trying to use libav to do some audio decoding, but unfortunately I am running into some issues
[10:46] <tetracon> everything seems to be fine until I try to assign a value to an int
[10:46] <tetracon> strange enough
[10:47] <tetracon> really breaking my brains on this one so it would great to get some support
[10:49] <cbreak> what do you get? What do you expect?
[10:50] <tetracon> http://pastebin.com/XUu4NFr7
[10:50] <tetracon> this is the code snippet
[10:50] <tetracon> I get a segfault after assigning the value to the int...
[10:51] <cbreak> well, that's weird.
[10:51] <cbreak> the int is on the stack, and an integer literal
[10:51] <tetracon> yes that is what[ bugging me :(
[10:52] <tetracon> if I don't assign 0 to i than there is no problem
[10:52] <tetracon> this should just be possible right?
[10:52] <burek> and you think it's that easy to assign a non-null value to an int?! :)
[10:53] <tetracon> lol it should be right?
[10:54] <cbreak> tetracon: I can't imagine any situation where assigning a number to an int variable would cause a crash when you assign it, depending on the value
[10:55] <cbreak> I think your diagnosis is flawed
[10:56] <tetracon> alright that is possible, however when I comment out the assigment it does not crash
[10:57] <tetracon> atm there isn't any code after this
[10:59] <tetracon> http://pastebin.com/mdUvWRVH
[10:59] <tetracon> this is the entire function, in this form is does not print anything to the console and produces the segfault
[11:01] <tetracon> with int i = 0; commented out it'll print this to the console: http://pastebin.com/0DYcsukV
[11:01] <cbreak> tetracon: if it depends on the value, then it doesn't crash because of the assignment
[11:01] <cbreak> it crashes because something you do because of the value in the variable later on
[11:02] <tetracon> I am not using the variable later on there is no more code
[11:02] <cbreak> nested comments?
[11:02] <cbreak> bad :/
[11:03] <tetracon> you're right I should not nest comments...
[11:03] <tetracon> but does it really make difference now?
[11:03] <cbreak> maybe.
[11:03] <cbreak> who knows what the compiler considers commented out
[11:08] <tetracon> Ok just to be sure I removed them: http://pastebin.com/uGv0KscK
[11:08] <tetracon> well the ones which weren't actually comments that is
[11:08] <tetracon> still the same issue :S
[11:09] <cbreak> and what's the issue?
[11:09] <cbreak> it crashes before you reach the breakpoint after Done?
[11:10] <tetracon> that's right, it does not reach Done
[11:12] <tetracon> and when not declaring/assigning i it works like it should
[11:12] <cbreak> how do you know?
[11:13] <tetracon> console output
[11:13] <cbreak> did you set a breakpoint at done?
[11:13] <cbreak> console output is worthless
[11:13] <cbreak> you don't even flush with std::endl
[11:13] <burek> tetracon, it is possible that your segfault happens before
[11:14] <burek> just when you hit the line with int, then it is triggered
[11:14] <cbreak> I think it happens after.
[11:14] <burek> could you try updating your gcc or something
[11:15] <tetracon> ok ok well I'll try that, but it is pretty strange though
[11:18] <burek> every time I had such issues, it was a bug in compiler
[11:18] <burek> some specific order of instructions
[11:18] <burek> that lead to a segf
[11:18] <burek> that's why I suggest you update your gcc
[11:21] <tetracon> ok ok I'll check if there are updates available
[11:21] <tetracon> thnx for the support so far :)
[11:21] <boys21> hi, I would like to add svc functionality to vlc player through ffmpeg
[11:21] <boys21> how can i add svcdecoder to ffmpeg
[11:21] <boys21> ?
[11:22] <boys21> gpac has svc decoder support in extra-libs
[11:22] <boys21> can any one suggest me
[11:23] <burek> boys21, you want to write a decoder for ffmpeg?
[11:27] <cbreak> tetracon: so, did you set a breakpoint or add std::endl?
[11:27] <boys21> writing a decoder from the scratch level is dificult for my level of programming expertise
[11:28] <tetracon> no not yet I moved my code to the main() function just to be sure there aren't other things going on
[11:28] <boys21> there are already existing svc decoders written in C
[11:28] <boys21> one such is included in gpac
[11:29] <boys21> in any-case will that be possible to link ffmpeg and make custom compile vlc player to make it svc enabled decoder is of my interest
[11:29] <tetracon> The segfault seems to be produced by the call to avformat_open_input
[11:31] <tetracon> and the problem was: I did not declare the AVFormatContext pointer as a null pointer
[11:33] <cbreak> standard problem :/
[11:33] <burek> boys21, try asking in #ffmpeg-devel
[11:33] <boys21> okay. Thank you!
[11:34] <tetracon> yup sorry for wasting your time :8
[14:10] <Radex> hi, i migrated from 0.7 to 0.10 and i see that 0.10 not using my 8 core cpu - i'm testing conversion on libx264 pass1/2 - somebody know that issue? not sure how to debug it..
[14:10] <Radex> 0.10 using 1-2 cores max. in my case
[14:31] <TH0MAS_> Im looking for some advice Ive done alot of reading but I seem to be going around in circle I want to setup a private media server for myself that is compatible with xbmc
[14:31] <TH0MAS_> Is this something ffmpeg can do alone or would I need another software installed on my server,  the server is remote and I will be holding all my content there
[14:32] <TH0MAS_> As I travel alot this is the most ideal situation to have my media on a server and then play it from within XBMC
[15:37] <vulkanr> hello
[15:40] <vulkanr> i have 2 ac3 files. when concatenated - they play flawlessly. however, independently converting each one to mp3 and concatenating the result - would produce a noticeable playback gap...
[15:40] <vulkanr> same if i try aac (in a mpegts container)
[15:44] <vulkanr> the source has 2 channels so no downmixing required
[15:44] <vulkanr> would appreciate any help, links or directions... thx.
[16:11] <Olson> heyo :)  I updated to latest version of ffmpeg and just having a liiiitle issue with command I used to use for recording desktop.. it used to be like
[16:11] <Olson> ffmpeg -f alsa -ac 2 -i hw:1,0 -f x11grab -r 30 -s 1920x1080 -i :0.0 -acodec pcm_s16le -vcodec libx264 -vpre lossless_ultrafast -threads 1 output.mkv
[16:12] <Olson> i understand now -vpre I have to put -preset instead
[16:12] <Olson> but lossless_ultrafast  seems to give it problem
[16:13] <Mavrik> Olson: -preset ultrafast -crf 0
[16:13] <Mavrik> IIRC
[16:13] <Olson> ahh oky.. yes that seems to work at least
[16:13] <Olson> that's the same thing you think yeah?
[16:14] <Mavrik> -crf 0 should tell libx264 you want losless encode
[16:14] <Mavrik> (I think -qp 0 is the alternatice)
[16:14] <Mavrik> *alternative
[16:14] <Olson> kewl ok thanks man
[16:14] <Olson> just as a side note..
[16:15] <Olson> using that to record desktop, is there any way to optimise it to use at little processing power as possible, or is it really as good as it gets
[16:15] <Olson> it's not a real issue or anything i just wonder if it can be done more efficiently
[16:15] <Mavrik> Olson: basically, veryfast/ultrafast, "-threads 0" is pretty much the best you can do
[16:15] <Mavrik> if you want H.264 output
[16:16] <Olson> I think I do what the H.264
[16:16] <Olson> unless there is something.. I'm missing
[16:17] <Olson> it works fantastic you know, so no complaints :)
[16:19] <Olson> I suppose 30FPS might be overkill for screencasts..
[16:28] <Mavrik> ;)
[16:29] <Olson> Mavrik, sorry to be a pain.. can I see somewhere docs for what -qp 0  is and why it is alternative to -preset lossless -crf 0  I can't seem to find in  http://ffmpeg.org/ffmpeg.html
[16:30] <Mavrik> Olson: that's because it's libx264 behaviour, not ffmpeg behaviour
[16:30] <Mavrik> you'll have to look in x264 docs
[16:30] <Olson> ow.. yes :P
[16:30] <Olson> sorry . hehe
[16:31] <Olson> qp produces larger files than --crf for the same visual quality.
[16:31] <Olson> apparently
[16:31] <JEEBsv> only with >8bit
[16:31] <JEEBsv> where crf 0 != lossless
[16:32] <JEEBsv> qp 0 is lossless with 8bit as well as >8bit encoding
[16:32] <JEEBsv> (in most cases your libx264 is 8bit so they're exactly the same -- in other words, lossless)
[16:32] <Olson> soo....
[16:32] <JEEBsv> crf 0 with 8bit libx264 is a shortcut for qp 0, they do the same thing
[16:32] <Olson> mmm.. im not too smart on bitrates and such to be honest
[16:32] <JEEBsv> they're not bitrate based either
[16:32] <JEEBsv> they're just encoding it lossless
[16:33] <Olson> would one be more "efficient" than the other in general terms you think.. for if i'm recording screen and trying to use as little processor as possible
[16:33] <JEEBsv> no, qp 0 and crf 0 are exactly the same with 8bit libx26
[16:33] <JEEBsv> *libx264
[16:34] <JEEBsv> the speed/compression is set by your preset
[16:34] <JEEBsv> or well, the preset you set
[16:34] <Olson> right you are :)
[16:34] <Olson> thanks very much
[16:38] <TH0MAS_> Anyone use Wowza in here/
[16:47] <Olson> good morrow kind sirs!   o/
[17:05] <westwind> there're question about the video decoding in windows, who can tell me whether or no the accelerator of graphic card is used for high definition video decoding
[17:40] <Mavrik> westwind: it isn't used in ffmpeg
[17:44] <Illarion> hello!
[17:45] <Illarion> when I use several inputs, one for video and other for audio, the resulting video file has length of longer input. Is it possible to a) truncate audio if it is longer or b) repeat audio if it is shorter ?
[17:53] <sacarasc> -loop_input -shortest
[17:53] <sacarasc> I think they're the two options you want...
[17:55] <Illarion> man says "Loop over the input stream. Currently it works only for image streams."
[20:01] <sgfgdf> hi, guys! is it possible to easily copy ID3 tags from a mp3 file to a aac file?
[20:49] <arow> can anyone give me an example of what a command to capture audio and video from separate devices (e.g. alsa and v4l2) at the same time would look like?
[20:51] <arow> this is an example of what i've tried and the result: http://pastebin.com/rA5uB9jp
[22:31] <burek> arow, instead of avplay, use ffmpeg
[22:32] <burek> like this ffmpeg -f video4linux2 -i /dev/video0 -f alsa -i hw:0 out.flv
[22:32] <arow> i thought avplay was essentially a renamed version of ffplay?
[22:32] <burek> hm, I really wouldn't know, never used it :/
[22:33] <aphid> i think it's libav's
[22:34] <arow> http://lwn.net/Articles/433348/
[22:34] <arow> ffmpeg is supposed to be transitioning to a naming scheme centered on libav, it seems
[22:34] <ubitux> it's not renamed
[22:34] <ubitux> it's a fork
[22:35] <ubitux> they can do whatever they want, half the dev left and created libav
[22:35] <arow> i've asked at libav too and gotten no response
[22:35] <ubitux> ffmpeg is still maintained by the other half
[22:35] <burek> arow, no, I think there more than that, something like, some guys wanted to make things in a different way so they forked ffmpeg and named it libav
[22:35] <arow> so ffmpeg is still maintained separately then?
[22:35] <burek> other guys just continued developing ffmpeg
[22:35] <ubitux> yes, and we merge everything from libav
[22:36] <ubitux> plus a lot of extra fix and additionnal features
[22:36] <ubitux> (which they don't pick most of the time)
[22:36] <ubitux> libav has av* tools, ffmpeg has ff* tools
[22:38] <arow> ah, the ffmpeg package in the ubuntu repos seems to just be there for compatibility as they move over solely to libav
[22:39] <arow> and the binaries are just renamed from libav
[22:40] <cbreak> what's libav good for?
[22:40] <cbreak> besides being non-compiling code
[22:41] <arow> i dunno, apparently someone compiled it because i have the binary right here
[22:41] <cbreak> maybe they got lucky :)
[22:41] <cbreak> I wasted so much time trying to get ffmpeg from git working
[22:41] <cbreak> and all I got was some av* things
[22:41] <cbreak> that were merged in some commits back
[22:42] <cbreak> and the whole thing crashed for the few commits where I actually got anyting
[22:42] <cbreak> two hours wasted on bisecting...
[22:42] <cbreak> (at least everything was fine before that av stuff was merged :)
[22:42] <arow> anyway sorry for asking in the wrong place, but this whole thing is pretty confusing and there seems to be a lot of conflicting info around
[22:43] <cbreak> I wouldn't waste time with it if you can help it
[22:43] <cbreak> when I gave up and switched to 0.10.1 everything worked :)
[22:44] <cbreak> (or some version that looked similar)
[22:45] <burek> arow, you are best off getting latest git version of ffmpeg
[22:45] <burek> it compiles and works
[22:46] <arow> thanks anyway
[23:36] <_tibo_> Hi, I'm pretty new to FFmpeg... If I have a h.264 RGB and use am I going to get a RGB frame or YUV ?
[23:37] <_tibo_> and use av_read_frame() and avcodec_decode_video2()
[00:00] --- Sat May  5 2012


More information about the Ffmpeg-devel-irc mailing list