[Ffmpeg-devel-irc] ffmpeg-devel.log.20181117

burek burek021 at gmail.com
Sun Nov 18 03:05:03 EET 2018

[00:54:12 CET] <raidensnake> there's a bug trying to configure ffmpeg on the raspberry pi 3. I get the error smbclient not found.
[00:55:54 CET] <BtbN> Install it, or don't enable it?
[00:56:07 CET] <raidensnake> it is installed
[00:56:31 CET] <raidensnake> it's saying it can't find libsmbcvlient.h when it exists/
[00:56:35 CET] <iive> you don't really need it
[00:57:04 CET] <iive> it is just so you can access files from remote cift/windows network without mounting it.
[00:57:19 CET] <iive> cifs
[00:58:22 CET] <raidensnake> only thing is it's needed for omxplayer
[01:09:32 CET] <iive> ?? why would you need samba for omxplayer??
[01:20:03 CET] <atomnuker> how is omxplayer 1.) not bitrotten to hell and 2.) not replaced by the crap but somewhat standard v4l2 hwaccel api already
[02:25:07 CET] <J_Darnley> too bad he left so I can't insinuate he's too dumb to read config.log
[11:14:33 CET] <cone-832> ffmpeg 03Paul B Mahol 07master:fbea2ea76ef9: avfilter/vf_neighbor: fix >8 bit deflate/inflate
[14:10:00 CET] <durandal_1707> libavcodec/libaomenc.c:730:9: warning: unused variable 'pict_type' [-Wunused-variable]
[17:16:54 CET] <durandal_1707> michaelni: see vf_blend code for grainextract, av_clip_uintp2 gives artifacts for bitdepth 10 & 12
[17:18:20 CET] <durandal_1707> either compiler miscompiles code or implementation of that function is faulty
[17:19:16 CET] <cone-561> ffmpeg 03Philip Langdale 07master:420ab946ace2: avutil/cuda_check: Make sure this passes make fate-source
[17:19:20 CET] <JEEB> would something like ubsan help with that?
[17:19:40 CET] <durandal_1707> i have clang 6.0.0 so that is probably reason...
[17:19:53 CET] <durandal_1707> lets see if i switch to gcc
[17:22:42 CET] <durandal_1707> same output, so it is broken code
[17:23:15 CET] <JEEB> no, ubsan is a module in some compilers which would find if something's UB (sometimes)
[17:23:23 CET] <JEEB> and it should warn about it
[17:28:01 CET] <durandal_1707> JEEB: right, it reports something
[17:28:11 CET] <JEEB> cool
[17:28:24 CET] <JEEB> hopefully it's useful :D
[17:46:17 CET] <BBB> atomnuker: have you made any progress on the EC SIMD?
[18:07:11 CET] <cone-561> ffmpeg 03Martin Vignali 07master:d3621b232142: avfilter: use av_clip_uintp2 instead of av_clip for 10b and 12b
[18:53:51 CET] <durandal_1707> it is too cold to write SIMD
[20:32:43 CET] <atomnuker> BBB: no, I was waiting for you to bug me again to get motivated to do it
[20:46:02 CET] <BBB> atomnuker: hm.. j-b is a better bugger than me, shall I ask him to join?
[20:48:42 CET] <durandal_1707> does motion compesation filter would just use motion vectors and copy block from previous or forward frame (whatever is uses forward or backward motion vectors)?
[20:51:50 CET] <kierank> BBB: that sentence means something rude/funny in british english
[20:52:18 CET] <BBB> <- foreigner
[20:52:25 CET] <BBB> see, that's why you folks voted for brexit
[20:52:30 CET] <BBB> you just don't understand me :-p
[20:53:04 CET] <kierank> https://en.oxforddictionaries.com/definition/bugger
[20:54:02 CET] <BBB> hehe :)
[20:56:37 CET] <durandal_1707> what innocent could happen with three guys together...
[20:57:42 CET] <JEEB> bink2 decoder?
[21:00:31 CET] <durandal_1707> bink2 decoder is far from innocent
[21:42:39 CET] <durandal_1707> looks like nobody here understand motion compensation concept...
[21:43:53 CET] <atomnuker> durandal_1707: yes, motion compensation is essentially just that - copy blocks of pixels from previous frames
[21:45:07 CET] <atomnuker> kierank: @the second definition of bugger - when did the oxford dictionary turn into the urbandictionary?
[21:45:29 CET] <kierank> it's a widely used insult
[21:45:35 CET] <kierank> not just a new one
[21:45:43 CET] <nevcairiel> just have enough people use something for long enough, and it gets written into dictionaries
[21:45:50 CET] <nevcairiel> language is fluid, afterall
[21:45:52 CET] <atomnuker> I know, didn't know about that definition of it though
[21:46:01 CET] <atomnuker> apparently its origin is australian
[21:46:40 CET] <nevcairiel> not being british, i only knew the first definition (ie. 1.1)
[22:13:15 CET] <rcombs> what's the difference between NVDEC and CUVID anyway
[22:16:44 CET] <nevcairiel> just how its implemented
[22:17:12 CET] <nevcairiel> nvdec is a proper hwaccel, so it gives you all proper metadata that the normal encoders would also give you, and cuvid is a decoder which probably lacks most of that handling =p
[22:17:38 CET] <rcombs> ahh
[22:26:39 CET] <philipl> BtbN: https://github.com/philipl/mpv/tree/cuda-semaphore-buffers
[22:26:44 CET] <philipl> Could you test that on windows for me?
[22:28:25 CET] <philipl> rcombs: cuvid also has worse compatibility because it relies on nvidia's decoder library which isn't as comprehensive as the ffmpeg ones; unsurprisingly.
[22:30:38 CET] <rcombs> mmmmmm
[22:30:46 CET] <rcombs> so the naming is just a bit misleading
[22:31:14 CET] <JEEB> first one got named cuvid, basically :P
[22:32:24 CET] <rcombs> I like the name "nvdec" a bit better if only for the symmetry with nvenc
[22:34:03 CET] <philipl> Originally nvidia called it cuvid, so we called the first impl cuvid too.
[22:34:10 CET] <philipl> Then they renamed it nvdec, so we used that for the hwaccel.
[22:34:21 CET] <philipl> It conveniently reflects the chronology.
[22:34:47 CET] <philipl> Old busted name for old busted decoder. :-)
[00:00:00 CET] --- Sun Nov 18 2018

More information about the Ffmpeg-devel-irc mailing list