[Ffmpeg-devel-irc] ffmpeg-devel.log.20181228

burek burek021 at gmail.com
Sat Dec 29 03:05:03 EET 2018


[00:22:58 CET] <cone-847> ffmpeg 03Carl Eugen Hoyos 07master:17aed996bc35: lavf/vividas: Do not increase extradata_size after allocation.
[00:22:59 CET] <cone-847> ffmpeg 03Carl Eugen Hoyos 07master:01db03f15848: lavf/vividas: Support demuxing on big-endian hardware.
[00:23:00 CET] <cone-847> ffmpeg 03Carl Eugen Hoyos 07master:1a6cca19898c: configure: Make sure libpostproc can be found if -rpath-link doesn't work.
[01:04:13 CET] <jamrial> BBB: can you take a look at the av1 frame split patch whenever you have time?
[01:04:24 CET] <jamrial> you don't need to understand how cbs works internally, just know that you pass it a packet and it returns an array of obus with filled values plus size and data pointers to the start of the revelant obu
[01:05:19 CET] <jamrial> much like in the vp9 superframe split, i'm using those values to create packet references pointing to different offsets in the buffer
[01:10:15 CET] <BBB> ok
[01:10:19 CET] <BBB> so it's obu splitter basically?
[01:11:47 CET] <jamrial> not into individual obus, into groups of obus that form a frame
[01:12:08 CET] <jamrial> so it could be a frame obu, or a frame header obu + one or more tile group obus
[01:14:33 CET] <jamrial> seq header is included with the first frame if present, any unrelated obu is included if present in between related obus (like padding or whatever in between frame header and tile groups and weird things like that), and any unrelated trailing obu in included with the last frame
[01:15:28 CET] <jamrial> i could use the writing features from cbs to actually create entire new packets with cherry picked obus, but that's slow and this bsf is meant to be used with a decoder, much like the vp9 one
[01:17:43 CET] <atomnuker> what's it meant to be fed by? shouldn't containers take care of that?
[01:18:32 CET] <BtbN> Well, users tend to do weird things
[01:18:36 CET] <BtbN> like, not using containers
[01:20:19 CET] <jamrial> atomnuker: you feed it a temporal unit, which can contain several frames
[01:21:03 CET] <jamrial> it's only useful for a decoder for frame multi threading purposes
[01:23:55 CET] <atomnuker> ah, I see
[11:41:48 CET] <durandal_1707> in all internet there is single ambisonic download, ignoring youtube untagged crap
[12:27:01 CET] <cone-293> ffmpeg 03Paul B Mahol 07master:cfa7709d05b1: avcodec/wavpack: fix decoding of files with many channels
[12:55:24 CET] <cone-293> ffmpeg 03David Bryant 07master:db109373d87b: avformat/wvdec: detect and error out on WavPack DSD files
[14:05:51 CET] <cone-293> ffmpeg 03Carl Eugen Hoyos 07master:8d422e7fef97: Makefile: Allow "make clean" to delete compat/atomics/pthread/stdatomic.o
[00:00:00 CET] --- Sat Dec 29 2018


More information about the Ffmpeg-devel-irc mailing list