[Ffmpeg-devel-irc] ffmpeg-devel.log.20161224

burek burek021 at gmail.com
Sun Dec 25 03:05:02 EET 2016


[02:12:33 CET] <atomnuker> jamrial: why does it require -strict -1?
[02:12:41 CET] <atomnuker> some container doesn't support it?
[02:13:02 CET] <jamrial> some container? this is about matroska specifically
[02:13:53 CET] <jamrial> it requires strict unofficial because when the colour elements were added to the muxer they were not yet made official in the spec
[02:14:06 CET] <jamrial> they are now, so i guess the check can be removed
[04:48:22 CET] <cone-877> ffmpeg 03James Almer 07master:94eb600f354c: tests/audio_fifo: fix buffer allocation for non planar formats
[10:41:40 CET] <cone-954> ffmpeg 03Paul B Mahol 07master:43cd33be16b2: avcodec/pixlet: fix undefined behaviour in postprocess_chroma
[10:41:40 CET] <cone-954> ffmpeg 03Paul B Mahol 07master:ab31b46b8936: avcodec/pixlet: avoid some overflows
[10:41:40 CET] <cone-954> ffmpeg 03Paul B Mahol 07master:6cedd20b9729: avcodec/pixlet: make sure scaling factors are not zero
[11:03:10 CET] <cone-954> ffmpeg 03Paul B Mahol 07master:68e5598e22b6: avcodec/utvideo: fix mistake using wrong arguments for left and lefttop pixel components
[14:26:16 CET] <cone-454> ffmpeg 03Nicolas George 07master:3b64e3ea45c5: lavf/framesync: detect EOF immediately.
[14:26:16 CET] <cone-454> ffmpeg 03Nicolas George 07master:0ff5567a30be: lavfi/buffersrc: push the frame deeper if requested.
[14:47:25 CET] <cone-454> ffmpeg 03Michael Niedermayer 07master:87f6f1546098: tools/zmqsend: Do not truncate fgetc() return
[14:47:26 CET] <cone-454> ffmpeg 03Michael Niedermayer 07master:c4152fc42e48: avutil/random_seed: Reduce the time needed on systems with very low precission clock()
[16:03:12 CET] <kierank> durandal_1707: "Apparently, and someone is making money from it."???
[16:03:53 CET] <durandal_1707> kierank: selling magicyuv encoder
[16:04:11 CET] <durandal_1707> it does 4k in real time 
[16:05:17 CET] <durandal_1707> and they claim their decoder is 2x faster than ours
[16:06:28 CET] <durandal_1707> and it's implementation improved utvideo basically
[16:49:02 CET] <cone-454> ffmpeg 03Paul B Mahol 07master:25c4035529c8: avcodec/pixlet: simplify lowpass_prediction() function
[17:46:15 CET] <kierank> durandal_1707: do you understand extradata parsing code
[17:47:42 CET] <kierank> mpeg-4 parser reads headers, puts them in extradata for some reason
[17:48:30 CET] <durandal_1707> what's problem?
[17:50:31 CET] <kierank> causes problems with the startcode detection I have written
[17:51:03 CET] <kierank> means I have to do all sorts of insane hacks
[17:52:57 CET] <durandal_1707> is extradata actually used?
[17:54:03 CET] <kierank> no it decides to put the data in extradata for some reason
[17:54:06 CET] <kierank> it's the frame headers
[18:59:02 CET] <Chloe> Merry Christmas All :)
[20:00:11 CET] <kierank> The other big question I have is how to get 32-bit coefficients into mpegvideo.c
[20:56:11 CET] <durandal_1707> kierank: via extradata or sidedata
[21:05:22 CET] <kierank> Huh?
[21:12:32 CET] <durandal_1707> kierank: your last question?
[21:20:11 CET] <kierank> durandal_1707: i need 32-bit idct coefficients for 10-bit mpeg-4 but mpegvideo is "designed" around 8-bit
[21:20:16 CET] <kierank> so I don't know how it make it use the right coefficients
[21:20:59 CET] <durandal_1707> rip you own version?
[21:23:21 CET] <Compn> cant you just add a new function in mpegvideo for 10bit ?
[21:23:27 CET] <Compn> duplicate where needed
[21:24:07 CET] <Compn> probably adding 10bit would slow down 8bit decoding anyway
[21:24:16 CET] <Compn> so might as well duplicate it and seperate path/decoder 
[21:24:25 CET] <kierank> michaelni: ping, how should I get 32-bit coeffiecients into mpegvideo
[21:28:45 CET] <kierank> would if(10bit) s->decode_mb10 work?
[21:30:17 CET] <JEEB> durandal_1707: seems like zimg will switch behavior without API changes with 2.4
[21:31:28 CET] <durandal_1707> JEEB: what exactly ?
[22:19:06 CET] <michaelni> kierank, Annex A says that the IDCT input is represented by n+4 bits, that would be 14 in case of 10bit and should not require 32bit
[22:20:14 CET] <kierank> hmmmmmmm
[22:20:26 CET] <kierank> do ffmpeg idcts support that range?
[22:20:57 CET] <michaelni> the SIMD optimized ones that generally output 8bit probably not
[22:21:32 CET] <kierank> prores one should work
[00:00:00 CET] --- Sun Dec 25 2016


More information about the Ffmpeg-devel-irc mailing list