[Ffmpeg-devel-irc] ffmpeg-devel.log.20160830

burek burek021 at gmail.com
Wed Aug 31 03:05:02 EEST 2016


[00:07:48 CEST] <BBB> what is p010be/le?
[00:08:20 CEST] <BBB> I remember cehoyos wanting something like that a while ago also (by setting bits_per_sample or so?) using a different pixfmt
[00:14:00 CEST] <jamrial> nv12 but 10bit, basically. for hwaccel
[00:18:33 CEST] <TD-Linux> bofh_, it's a lossless format, likely an attempt to avoid bit-inexactness of float implementations
[00:21:28 CEST] <iive> 10 most significant bits
[00:41:55 CEST] <cone-427> ffmpeg 03Stephen Hutchinson 07master:92916e8542e4: avisynth: support pix_fmts added to AviSynth+
[00:57:52 CEST] <durandal_1707> does rust have templates?
[00:59:23 CEST] <TD-Linux> it has generics, so "yes"
[02:01:05 CEST] <Plorkyeran> and it has hygenic macros
[02:56:55 CEST] <cone-427> ffmpeg 03James Almer 07master:115a90a0dbeb: doc/APIChanges: mention nb_threads addition to AVFilterContext
[04:09:48 CEST] <cone-427> ffmpeg 03Tobias Rapp 07master:a64839189622: cmdutils: fix implicit declaration of SetDllDirectory function
[04:09:51 CEST] <cone-427> ffmpeg 03Tobias Rapp 07release/2.8:c32ce247a026: cmdutils: fix implicit declaration of SetDllDirectory function
[04:09:54 CEST] <cone-427> ffmpeg 03Tobias Rapp 07release/3.0:426b959e37bd: cmdutils: fix implicit declaration of SetDllDirectory function
[04:09:58 CEST] <cone-427> ffmpeg 03Tobias Rapp 07release/3.1:12320c08221f: cmdutils: fix implicit declaration of SetDllDirectory function
[10:14:17 CEST] <BtbN> Compn, no, slow as in "it takes an hour for my own mails to reach me via the list"
[10:36:39 CEST] <durandal_1707> michaelni: scale filter still doesn't use slice threading?
[10:46:44 CEST] <michaelni> no i think noone implemeted that yet
[10:47:21 CEST] <durandal_1707> michaelni: but swscale supports it?
[10:47:41 CEST] <durandal_1707> I wonder how was it tested?
[10:51:38 CEST] <michaelni> swscale supports slices but theres no native thread support, so one could do slice threads for unscaled convertions from the outside (with multiple contexts), for scaled cases changes in sws are needed unless ive forgotten about some change
[10:52:49 CEST] <michaelni> also talk with Pedro Arthur, he worked on code related to this
[14:18:17 CEST] <BtbN> 1h15min for this mail to get through, hm
[15:11:43 CEST] <ubitux> i'm confused
[15:11:52 CEST] <ubitux> what's the difference between av_log_format_line() and av_log_format_line2()?
[15:14:03 CEST] <JEEB> o_O
[15:14:17 CEST] <JEEB> oh right, that's different level from av_log
[15:18:40 CEST] <durandal_1707> michaelni: i plan to do frame mt with caches of size bb_threads, for each filter, this eats RAM
[16:03:10 CEST] <BtbN> hm, something feels broken with the ML. It's taking WAY too long to deliver mails.
[16:06:20 CEST] <BtbN> https://bpaste.net/show/05cc5caaaa3f something on the ffbox0-bg.mplayerhq.hu is taking its time
[16:06:52 CEST] <BtbN> michaelni, ^ or whoever is in charge of those boxes?
[17:20:02 CEST] <cone-287> ffmpeg 03Simon Hailes 07master:a2fcacc880ca: libavformat/crypto - encourage reads of 4096 bytes
[17:20:03 CEST] <cone-287> ffmpeg 03Simon Hailes 07master:a13a81a0dc98: avformat/crypto: add seeking support on read
[18:29:19 CEST] <michaelni> BtbN, ML should be fast again, i hope
[19:40:49 CEST] <BtbN> what was the issue?
[19:52:34 CEST] <michaelni> mailman+postfix timeouts+dns failures causing longer than normal delays
[20:35:47 CEST] <BtbN> The interest on the P010 sws output seems to be incredibly high.
[20:43:17 CEST] <fritsch> hehe
[23:36:04 CEST] <kierank> is it possible to pin threads to certain cores in ffmpeg
[23:36:15 CEST] <kierank> i.e pin the downstream threads that are created by lavc
[23:36:24 CEST] <kierank> without doing weird pidof trickery
[23:48:37 CEST] <Compn> on linux? bsd ?
[23:48:39 CEST] <Compn> x64 ? 
[23:59:28 CEST] <DHE> kierank: ffmpeg doesn't do it internally, you'll have to do something external. under linux you might use control groups or the `taskset` utility
[00:00:00 CEST] --- Wed Aug 31 2016


More information about the Ffmpeg-devel-irc mailing list