[Ffmpeg-devel-irc] ffmpeg-devel.log.20111110

burek burek021 at gmail.com
Fri Nov 11 02:05:03 CET 2011


[01:42] <CIA-18> ffmpeg: 03Mark Himsley 07master * r0d602615ab 10ffmpeg/ (3 files in 2 dirs): 
[01:42] <CIA-18> ffmpeg: vf_fade: add support to options parsing
[01:42] <CIA-18> ffmpeg: Extend fade syntax, required by a pending patch.
[01:42] <CIA-18> ffmpeg: Signed-off-by: Stefano Sabatini <stefasab at gmail.com>
[01:42] <CIA-18> ffmpeg: 03Mark Himsley 07master * r8c1fb50d07 10ffmpeg/ (3 files in 2 dirs): 
[01:42] <CIA-18> ffmpeg: vf_fade: add alpha option and alpha fading support
[01:42] <CIA-18> ffmpeg: Signed-off-by: Stefano Sabatini <stefasab at gmail.com>
[01:42] <CIA-18> ffmpeg: 03Stefano Sabatini 07master * rdec354ba1d 10ffmpeg/libavfilter/vf_fade.c: vf_fade: reindent after previous commit
[03:22] <ohsix> is there something that can just change the quant on an mpeg4 stream for a given set of frames; i goofed and i want to take some already encoded files, change the quant on the frames i should have initially; see how much more bitrate it gets me for the size
[03:59] <CIA-18> ffmpeg: 03shahriman AMS 07master * r6475a6e1ca 10ffmpeg/libavcodec/vc1dec.c: 
[03:59] <CIA-18> ffmpeg: vc1dec: take ME precision into account while scaling MV predictors.
[03:59] <CIA-18> ffmpeg: Signed-off-by: Anton Khirnov <anton at khirnov.net>
[03:59] <CIA-18> ffmpeg: 03shahriman AMS 07master * r4344ce08f1 10ffmpeg/libavcodec/vc1dec.c: (log message trimmed)
[03:59] <CIA-18> ffmpeg: vc1dec: Partial rewrite of vc1_pred_mv()
[03:59] <CIA-18> ffmpeg: The current code is a generalization of the earlier progressive MV
[03:59] <CIA-18> ffmpeg: prediction code. This was supposed to predict MVs for both interlaced
[03:59] <CIA-18> ffmpeg: and progressive pictures. But the interlaced MV prediction is buggy
[03:59] <CIA-18> ffmpeg: and works mostly by luck.
[03:59] <CIA-18> ffmpeg: This partially fixes interlaced MV prediction.
[03:59] <CIA-18> ffmpeg: 03shahriman AMS 07master * rc33d4916fb 10ffmpeg/libavcodec/vc1dec.c: 
[03:59] <CIA-18> ffmpeg: VC1: Add bottom field offset to block_index[] to avoid rewriting (+10L)
[03:59] <CIA-18> ffmpeg: Signed-off-by: Anton Khirnov <anton at khirnov.net>
[03:59] <CIA-18> ffmpeg: 03Martin Storsjö 07master * r65ac51373b 10ffmpeg/ (doc/protocols.texi libavformat/udp.c): 
[03:59] <CIA-18> ffmpeg: udp: Allow specifying the local IP address
[03:59] <CIA-18> ffmpeg: This is useful if sending multicast data on a host with
[03:59] <CIA-18> ffmpeg: multiple interfaces.
[03:59] <CIA-18> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[03:59] <CIA-18> ffmpeg: 03Justin Ruggles 07master * rf1f6d3615f 10ffmpeg/ (6 files in 2 dirs): 
[03:59] <CIA-18> ffmpeg: avcodec: add support for planar signed 8-bit PCM.
[03:59] <CIA-18> ffmpeg: It is found in some 8svx files (e.g. ones created by SoX).
[03:59] <CIA-18> ffmpeg: Currently the decoder reuses the 8svx functions because we already have
[03:59] <CIA-18> ffmpeg: handling of a single large planar packet for the compressed 8svx codecs.
[03:59] <CIA-18> ffmpeg: 03Andrey Utkin 07master * r055a141e44 10ffmpeg/libavformat/utils.c: 
[03:59] <CIA-18> ffmpeg: lavf: don't leak corrupted packets
[03:59] <CIA-18> ffmpeg: Signed-off-by: Anton Khirnov <anton at khirnov.net>
[03:59] <CIA-18> ffmpeg: 03shahriman AMS 07master * r4bceeaf0c1 10ffmpeg/libavcodec/vc1dec.c: 
[03:59] <CIA-18> ffmpeg: vc1dec: use correct hybrid prediction threshold.
[04:00] <CIA-18> ffmpeg: Signed-off-by: Anton Khirnov <anton at khirnov.net>
[04:00] <CIA-18> ffmpeg: 03Justin Ruggles 07master * r818a3bdedf 10ffmpeg/libavcodec/smacker.c: 
[04:00] <CIA-18> ffmpeg: smackaud: use uint8_t* for 8-bit output buffer type
[04:00] <CIA-18> ffmpeg: This matches the output sample format and the range of the output values.
[04:00] <CIA-18> ffmpeg: 03Justin Ruggles 07master * rfed5ca255f 10ffmpeg/libavcodec/ra144enc.c: 
[04:00] <CIA-18> ffmpeg: ra144enc: add sample_fmts list to ff_ra_144_encoder
[04:00] <CIA-18> ffmpeg: This enables the user (i.e. avconv) to automatically convert to the supported
[04:00] <CIA-18> ffmpeg: sample format if needed.
[05:36] <CIA-18> ffmpeg: 03Michael Niedermayer 07master * r525a8f3ba5 10ffmpeg/libavcodec/avcodec.h: 
[05:36] <CIA-18> ffmpeg: lavc: update AVCodecContext.delay documentation
[05:36] <CIA-18> ffmpeg: Based on reimars suggestion
[05:36] <CIA-18> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[05:36] <CIA-18> ffmpeg: 03Michael Niedermayer 07master * r3cc1af9dd9 10ffmpeg/libavcodec/ (pthread.c version.h): 
[05:36] <CIA-18> ffmpeg: pthreads: export the delay that was previously exported by mangling has_b_frames through
[05:36] <CIA-18> ffmpeg: AVCodecContext.delay which was previously unused for decoders and fits better.
[05:36] <CIA-18> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[15:38] <CIA-18> ffmpeg: 03Michael Niedermayer 07master * reb4fc6acfe 10ffmpeg/ffmpeg.c: 
[15:38] <CIA-18> ffmpeg: ffmpeg: update timestamping code to handle frame threads
[15:38] <CIA-18> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[15:38] <CIA-18> ffmpeg: 03Michael Niedermayer 07master * r80f716cd45 10ffmpeg/libavcodec/mpegvideo.c: 
[15:38] <CIA-18> ffmpeg: mpeg4videodec: fix lml4 and frame threading without -vsync 0
[15:38] <CIA-18> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:52] <pasteeater> michaelni: what do you need from me so i can make trivial updates to the site (such as removing that superfluous | in navbar)?
[21:55] <michaelni> public ssh key, gpg signed
[21:56] <michaelni> and a patch with the gpg key fingerprint for the MAINTAINERS file
[21:57] <pasteeater> ok. thanks.
[21:58] <cbsrobot> michaelni: did you add something like this to trac ?  http://blog.timp.com.au/archives/2010/08/25/speedup_of_trac_static_content/
[22:12] <Daemon404> michaelni, for the life of me, i cannot get add_hfyu_median_prediction to work like you said in your email
[22:12] <Daemon404> (replacing the code i have now)
[22:15] <fritsch> hi and good evening a nice helping hand got me from ffmpeg to here
[22:15] <Daemon404> er wait.. :3
[22:15] <fritsch> a collegue and me are developing direct xvba support for ffmpeg
[22:15] <fritsch> to have it in xbmc
[22:16] <fritsch> i have some questions concerning vc1
[22:16] <fritsch> is it okay to speak up?
[22:16] <Compn> yes
[22:16] <Compn> fritsch : go ahead
[22:17] <fritsch> https://github.com/fritsch/xbmc/blob/28cbc48a4e8df0a380dc08f10f094f9347326318/lib/ffmpeg/libavcodec/xvba_vc1.c
[22:17] <fritsch> this is currently the code we use for struct filling
[22:17] <fritsch> the xvba structures
[22:17] <fritsch> i could not find anything obiously wrong in it
[22:18] <fritsch> but all i get is totally dark green picture
[22:18] <fritsch> someone mentioned to get coded_width and height from mpeg context, i am trying this now
[22:18] <nevcairiel> not related to your question, but to your project; can't you use xvba through vaapi?
[22:19] <fritsch> this was our first try ...
[22:19] <fritsch> we wanted - to have one abi
[22:19] <fritsch> but xvba-driver did not get updated since june
[22:20] <nevcairiel> i see
[22:20] <fritsch> and there is so much gl stuff in xvba-driver that we never could get it to "clean up" correctly
[22:20] <nevcairiel> so the xvba vaapi interface just sucks, eh :)
[22:20] <fritsch> i don`t want to say it like this
[22:20] <fritsch> i really don`t want to
[22:21] <fritsch> it was a good job done - and many fglrx users were happy
[22:21] <fritsch> but we have long lists of bugs and forum entries of users which do not get it to work correctly
[22:21] <nevcairiel> looking at your code - why do you fill the structs in end_frame and not in start_frame?
[22:22] <fritsch> i looked like the others are doing it
[22:22] <fritsch> vaapi, vdpau
[22:22] <fritsch> i am really new to all this stuff - the real answer is a char* which is "don`t really know" :-)
[22:23] <fritsch> i take every information you will give me, write it down and will read it up afterwards
[22:24] <nevcairiel> if i look at vaapi or dxva, they all seem to do it in start frame
[22:24] <nevcairiel> which in my mind also makes more sense
[22:24] <nevcairiel> You want the infos before you start decoding. :)
[22:25] <fritsch> this is correct :-)
[22:26] <nevcairiel> vdpau seems to be odd, it does it in end frame as well
[22:27] <fritsch> i don`t think that it really matters (from my point of view) cause only finished frames are send to decode
[22:27] <fritsch> meaning fully loaded structs
[22:28] <nevcairiel> does your h264 work?
[22:28] <fritsch> it does
[22:28] <fritsch> we just have some problems with filling the buffer, but it works - we get a picture
[22:28] <fritsch> but the code is far from finished ... somewhere a memcpy goes wrong and we get sever problems ... but alpha state
[22:29] <fritsch> working at this since nearly a week with a collegue
[22:29] <fritsch> is there a "this could be it" answer, if you just get a "dark green" image?
[22:29] <fritsch> something like iq matrix wrong in h264 or interlace flag wrong?
[22:29] <fritsch> just in the vc1 sense
[22:30] <nevcairiel> Is the vc1 interlaced? if so, try with progressive. :)
[22:31] <fritsch> you mean the coded video?
[22:31] <nevcairiel> yes
[22:31] <fritsch> are there any sources for this kind of material? i found something on the mplayer sites
[22:33] <fritsch> i try with progressive content, btw.
[22:33] <fritsch> but will look at the flag
[22:35] <michaelni> dark green sounds like luma & chroma values being too low
[22:35] <michaelni> why that is so, dont ask me ...
[22:35] <Daemon404> michaelni, yeah i  can't figure out to make add_hfyu_median_prediction work the way i need in vble, or if it is even possible.
[22:35] Action: Daemon404 has given up on that fr now.
[22:36] <michaelni> well, skip it then and after its in git ping me and ill try
[22:36] <Daemon404> thanks
[22:36] <fritsch> michaelni: thx
[22:36] <nevcairiel> fritsch: one thing i noticed is that you seem to insist on decoding the marker/startcode in decode_slice, while the other hardware decoders seem to prefer to skip it
[22:37] <Compn> dark green sounds like wrong colorspace :P
[22:37] <michaelni> or 9 vs 10 bit messup but it can be many other things
[22:37] <Compn> right
[22:37] <Compn> somethigns wrong :)
[22:37] <fritsch> oki, i write everything down
[22:37] <nevcairiel> I doubt its 9 vs 10bit on vc1 ;)
[22:38] <fritsch> first what  nevcairiel mentioned could be one source
[22:38] <fritsch> stept through the debugger and v->interlace says 1
[22:38] <nevcairiel> that doesn't mean much yet
[22:38] <michaelni> is the video decoded ok and just dar green or somehow ztrashed/ full of artifacts ?
[22:38] <fritsch> dark green
[22:38] <fritsch> not artefacts
[22:39] <michaelni> colors ? or just monochrome ?
[22:39] <fritsch> i will make you a photo of the screen
[22:39] <fritsch> :-)
[22:39] <nevcairiel> so you can see the image, its just green? Or its completely green, no shapes and forms?
[22:39] <fritsch> noe image
[22:39] <fritsch> totally green
[22:40] <fritsch> but in correct ratio :-)
[22:40] <nevcairiel> I would check the startcode thing, its the biggest difference i could find between the different decoders
[22:40] <nevcairiel> anyhow, time for sleep for me
[22:40] <fritsch> thank you very much
[22:42] <fritsch> thanks again, i will report back, if it works
[22:42] <Compn> huh that looks like spam
[23:00] <pasteeater> Compn: what looks like spam?
[23:01] <Compn> 16:43] * FARE-SEARCH_COM (~gfdffdfd at pool-74-111-197-200.lsanca.fios.verizon.net) Quit (Ping timeout: 256 seconds)
[23:02] <pasteeater> ah. i disabled those messages.
[23:09] <CIA-18> ffmpeg: 03Carl Eugen Hoyos 07master * r0e609d7418 10ffmpeg/libavcodec/bmp.c: 
[23:09] <CIA-18> ffmpeg: Fix some 1bit bmp samples, they are palletised.
[23:09] <CIA-18> ffmpeg: Fixes ticket #632.
[23:33] <CIA-18> ffmpeg: 03Lou Logan 07master * ra14708d4ac 10ffmpeg/doc/ffmpeg.texi: 
[23:33] <CIA-18> ffmpeg: doc: update examples to use -b:v or -b:a
[23:33] <CIA-18> ffmpeg: Fixes Ticket613
[23:33] <CIA-18> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:33] <CIA-18> ffmpeg: 03Nir Ben David 07master * r083d9ba448 10ffmpeg/libavcodec/golomb.h: 
[23:33] <CIA-18> ffmpeg: golomb: Fix overread in get_ur_golomb_jpegls()
[23:33] <CIA-18> ffmpeg: Fixes Ticket600
[23:33] <CIA-18> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:00] --- Fri Nov 11 2011


More information about the Ffmpeg-devel-irc mailing list