[Ffmpeg-devel-irc] ffmpeg-devel.log.20150405

burek burek021 at gmail.com
Mon Apr 6 02:05:02 CEST 2015


[00:53:39 CEST] <cone-586> ffmpeg 03Vignesh Venkatasubramanian 07master:2171b7c58e83: matroskadec: Add support for parsing live header files
[00:57:47 CEST] Action: Compn reads epic gentoo forum thread https://forums.gentoo.org/viewtopic-t-1010096-postdays-0-postorder-asc-start-50.html
[01:02:53 CEST] <Daemon404> that is pretty overwhelming
[01:02:59 CEST] <Daemon404> but unsurprising
[01:03:29 CEST] <Daemon404> also people still use forums?
[01:03:32 CEST] <Daemon404> what year is it?
[01:03:40 CEST] <Daemon404> (says the man on irc)
[01:03:53 CEST] <Compn> what, you prefer facebook or twitter ?
[01:03:56 CEST] <Compn> pick one, its 2015.
[01:04:11 CEST] <rcombs> don't forget tumblr
[01:05:04 CEST] <Daemon404> let me rephrase
[01:05:07 CEST] <Daemon404> the forums look like 1998.
[01:05:15 CEST] <Daemon404> (probably havent been updated since)
[01:05:45 CEST] <cone-586> ffmpeg 03Himangi Saraogi 07master:aa1a1b2496cf: avconv: Avoid theoretical NULL dereferences
[01:05:46 CEST] <cone-586> ffmpeg 03Michael Niedermayer 07master:bc0e65e7d0d0: Merge commit 'aa1a1b2496cf3ca817e78e27bd0262a50adb91a1'
[01:09:09 CEST] <Compn> Daemon404 : you want rotating html5 dancing forum posts? or discus forums ? 2015! whoooo
[01:09:56 CEST] <Daemon404> id take that over horrible purple and a logo made in gimp
[01:17:18 CEST] <rcombs> "modern web" is just a color scheme and some rounded corners, yeah?
[01:34:28 CEST] <jamrial> Daemon404: their main website looked like straight out of 1998 as well, but they upgraded it a couple days ago
[01:37:22 CEST] <Daemon404> oh yeah it looks way less getto now
[02:10:29 CEST] <cone-586> ffmpeg 03Vignesh Venkatasubramanian 07master:26f2e2f3f73f: webmdashenc: Support for live stream manifests
[02:28:52 CEST] <Compn> rcombs : brushed metal and white plastic
[02:29:06 CEST] <rcombs> that's Apple
[02:29:18 CEST] <Compn> shh you'll make appl stock drop :p
[02:29:45 CEST] <rcombs> apparently it doesn't take a lot of effort to manipulate the stock market these days
[02:34:12 CEST] <Compn> michaelni : what about a patch to optionally rename our libs to be installable next to libav ? like --libprefix=ff would make fflibavcodec.so fflibavresample.so ... its not optimal or even a good idea, but if it will fix distro problems then its in our best interests to fix it , yeah ?
[02:34:46 CEST] <Compn> although all other programs would have to switch so ... broken.
[02:35:05 CEST] <Prelude2004c> hi everyone.. can anyone give me some input on : -c:v libx264 -preset $5 -r 30 -g 20 -keyint_min 90 -bf 0 -b:v ${bitrate} -minrate 0k -maxrate ${maxrate} -bufsize 1000k -s ${resolution} \
[02:35:06 CEST] <Prelude2004c>         -flags +loop -me_method hex -g 250 -qcomp 0.6 -qmin 10 -qmax 51 -qdiff 4 -bf 16 -b_strategy 1 -i_qfactor 0.71 -cmp +chroma -subq 1 -me_range 16 
[02:35:06 CEST] <Prelude2004c> -coder 1 -sc_threshold 40  -keyint_min 25 -refs 1 -trellis 0 -partitions -parti8x8-parti4x4-partp8x8-partp4x4-partb8x8-an
[02:35:27 CEST] <Prelude2004c> do you guys thing that is good for the h265 output of 5Mbit/s or 3Mbit/s ?
[02:35:49 CEST] <Prelude2004c> i been playing with settings but i am seeing a few problems with fast movement where it looks like the images starts clipping a bit on fast movement parts
[02:36:19 CEST] <Prelude2004c> oh.. worng channel. .sorry was meant for FFMPEG
[02:37:11 CEST] <jamrial> Compn: isn't that what --build-suffix is for? both projects have that configure option, even
[02:37:30 CEST] <Compn> dont know never tried jamrial
[02:38:53 CEST] <jamrial> --build-suffix=suf turns the library name into libavcodec-suf.so
[02:40:01 CEST] <Compn> ah interesting
[02:40:15 CEST] <Compn> ok then , i was reading into uninformed ideas on a webforum
[02:40:18 CEST] <Compn> michaelni : nevermind
[04:24:46 CEST] <cone-586> ffmpeg 03Andreas Cadhalpun 07master:0224b35c89e9: avcodec: add unpack packed B-frames bitstream filter
[04:37:28 CEST] <aetasx> Daemon404: you still around?
[11:22:10 CEST] <markus__> hi, does anyone know about the mpeg2 video codec? I thought 00 00 01 e0 - ef contains the coded payload video but it seems like it's interrupted by other packets eg. 00 00 01 b2; does anyone know some good documentation or overview to understand the packet sequences?
[13:29:07 CEST] <cone-817> ffmpeg 03Stephan Holljes 07master:4a58d757de28: lavf/http: add proof-of-concept http server.
[14:01:20 CEST] <cone-817> ffmpeg 03Michael Niedermayer 07master:cf880ccb6a82: avcodec/error_resilience: Avoid race with updating the error_count
[14:35:23 CEST] <Compn> markus__ : it could be subtitles, metadata, or other info in the stream
[14:35:35 CEST] <Compn> there are some mpeg2 specs somewhere
[14:35:50 CEST] <Compn> markus__ : what container is your mpeg2 in ? ts? mpg ps? 
[14:50:07 CEST] <markus__> mpg ps
[15:30:51 CEST] <cone-817> ffmpeg 03Andreas Cadhalpun 07master:71fa0aa085cc: mpeg4: suggest mpeg4_unpack_bframes bsf instead of avidemux/VirtualDub
[16:07:05 CEST] <cone-817> ffmpeg 03Michael Niedermayer 07master:9caec04a3139: avfilter/vsrc_mandelbrot: Change enums to int, which are accessed via AVOption as int
[16:07:06 CEST] <cone-817> ffmpeg 03Michael Niedermayer 07master:0968180f028a: avfilter/yadif: Change enums to int, which are accessed via AVOption as int
[16:23:37 CEST] <cone-817> ffmpeg 03Anton Khirnov 07release/2.4:420aa06a2487: avconv: do not overwrite the stream codec context for streamcopy
[16:23:38 CEST] <cone-817> ffmpeg 03Michael Niedermayer 07release/2.4:17b27a7cbe26: ffmpeg: Fix extradata allocation
[16:23:39 CEST] <cone-817> ffmpeg 03Michael Niedermayer 07release/2.4:4347cf9f0fc7: Merge commit '420aa06a2487469259a04f9be66fd15535372796' into release/2.4
[16:27:50 CEST] <loki_> can somebody explain me how i can add cudart dep to ffmpeg build proccess?
[16:28:07 CEST] <loki_> i now found only one way threw configure option
[17:53:05 CEST] <Daemon404> loki_, all optional deps are configure options
[17:53:11 CEST] <Daemon404> aside fro nsystem-level libs like zlib.
[17:53:14 CEST] <Daemon404> from*
[18:26:06 CEST] <loki_> Daemon404: how i can add optional rule to build proccess?
[18:26:33 CEST] <loki_> i need build libavcodec/gpu_convert.cu with my command
[18:26:38 CEST] <loki_> how i can do this?
[18:28:43 CEST] <Daemon404> theyre just makefiles...
[18:31:24 CEST] <wm4> what is libavcodec/gpu_convert.cu?
[18:32:37 CEST] <loki_> it is my hack... little cuda kernel and functions which use it
[18:34:12 CEST] <wm4> ah, more nvidia specific code
[18:34:39 CEST] <loki_> yes
[18:34:45 CEST] <loki_> it must build using nvcc
[18:36:00 CEST] <wm4> make a generic rule for .cu (or whatever the source file extension is), and the rest will be like with .c
[18:37:33 CEST] <loki_> in build proccess i see many files like nvenc.d
[18:37:40 CEST] <loki_> or something else like *.d
[18:37:48 CEST] <loki_> who generates them?
[18:38:05 CEST] <loki_> it is like something dependencies as i understand
[18:39:40 CEST] <wm4> the compiler generates them because we pass -MD to it (or some similar flag)
[18:40:05 CEST] <wm4> the makefiles include these .d files (they are make rules), so changing a header will rebuild all .c files that include it
[18:43:13 CEST] <loki_> so where i need to write rules for gpu_convert.o? In liavcodec/MakeFile right?
[18:43:22 CEST] <loki_> libavcodec/MakeFile
[18:59:15 CEST] <BtbN> What do you want to do with CUDA?
[19:01:33 CEST] <loki_> i already done, it handles bgr0 to nv12 convertion
[19:01:54 CEST] <loki_> remember when we talk about this on #ffmpeg?
[19:02:08 CEST] <loki_> about that my proccessor lack support of ssse3
[19:02:13 CEST] <Daemon404> why would that be in avcodec at all
[19:02:15 CEST] <BtbN> I'm not sure if it's worth it to do that on the GPU
[19:02:24 CEST] <loki_> it worth
[19:02:27 CEST] <BtbN> And, avcodec is entirely the wrong place.
[19:02:31 CEST] <kierank> Daemon404: better in avcodec than swscale
[19:02:33 CEST] <kierank> =p
[19:02:42 CEST] <BtbN> It takes _a lot_ of bandwidth from the bus. Raw frames are huge.
[19:03:14 CEST] <loki_> i have now 30 stable fps
[19:03:22 CEST] <BtbN> The upload, download (and re-upload for nvenc) might be slower than software-convert with propper optimizations.
[19:04:32 CEST] <BtbN> Might be faster on horrible CPUs which lack basic instruction sets, but adding a CUDA dependency to ffmpeg just for those?
[19:05:36 CEST] <loki_> everybody have some problems, i think my is that i always took wrong way.
[19:05:53 CEST] <loki_> i just want to finish my little project.
[19:07:35 CEST] <BtbN> OpenCL might have a higher chance to be included.
[19:08:05 CEST] <loki_> you mean better to do this on opencl?
[19:08:12 CEST] <BtbN> yes
[19:08:30 CEST] <loki_> maybe... but i started learning with cuda.
[19:08:42 CEST] <BtbN> They aren't exactly very diffrent.
[19:08:57 CEST] <loki_> i will try.
[19:09:17 CEST] <BtbN> hardware accelerated color conversions are definitely interesting
[19:09:26 CEST] <BtbN> CUDA is possible the biggest showstopper, because nvidia-only.
[19:09:46 CEST] <loki_> i found it very user friendly.
[19:10:11 CEST] <BtbN> OpenCL is very similar, you most likely can keep most of your code.
[19:10:30 CEST] <BtbN> And it works on all GPUs
[19:24:10 CEST] <loki_> i found that in common.mak are rules for different .o am i right that i need add my rules for gpu_convert.o there?
[19:24:54 CEST] <loki_> i tried but see that my object file did not built again.
[19:26:07 CEST] <loki_> ok... i add there %.o: %.cu and i see that it tries to build my object file
[19:32:36 CEST] <wm4> <BtbN> The upload, download (and re-upload for nvenc) might be slower than software-convert with propper optimizations. <- but still faster than libswscale? trollol
[19:38:17 CEST] <loki_> wm4: thats why i passed libswscale and added bgr0 as supported format to nvenc and done cuda conversion there.
[19:52:22 CEST] <BtbN> wm4, on a AMD Phenom, without most modern extensions, propably.
[20:03:24 CEST] <BtbN> loki_, a bgr0 -> nv12 converter in libavcodec will be rejected though
[20:04:05 CEST] <loki_> i don't even plan post something.
[20:04:11 CEST] <loki_> i think all i done is bad.
[20:05:00 CEST] <loki_> will buy something which can be called as "proccessor" and will start use ffmpeg main.
[20:55:50 CEST] <cone-844> ffmpeg 03Anton Khirnov 07master:84f226a3bcd8: h264: use the correct SPS during PPS parsing
[20:55:50 CEST] <cone-844> ffmpeg 03Michael Niedermayer 07master:cde98920da78: Merge commit '84f226a3bcd8b39801a4c9051c033ab7d61aaf76'
[21:09:03 CEST] <cone-844> ffmpeg 03Anton Khirnov 07master:050324d020f8: h264: initialize cur_chroma_format_idc
[21:09:04 CEST] <cone-844> ffmpeg 03Michael Niedermayer 07master:68f48406069e: Merge commit '050324d020f843ce333276ebb6f27cc6026f37d0'
[21:11:30 CEST] <Daemon404> wut
[21:12:18 CEST] <kierank> ffmpeg supports 12-bit?
[21:12:38 CEST] <Daemon404> yes afaik
[21:13:06 CEST] Action: kierank isn't sure about that
[21:13:53 CEST] <Daemon404> i believe some fate tests use 12bit avc references
[21:14:02 CEST] <Daemon404> as well as 14 bit
[21:21:12 CEST] <kierank> I suspect they are wrong compared to jm
[21:29:46 CEST] <Daemon404> kierank, only one way to be sure
[21:29:54 CEST] <Daemon404> however since it is easter, im done poking code for day
[21:29:56 CEST] <Daemon404> today*
[21:34:53 CEST] <kierank> same
[21:46:48 CEST] <cone-844> ffmpeg 03Anton Khirnov 07master:43fd3dd80ca2: h264: drop redundant initialization in init()
[21:46:49 CEST] <cone-844> ffmpeg 03Michael Niedermayer 07master:14d241b19dcd: Merge commit '43fd3dd80ca2d1c2ccf6a7b7632db544c809c690'
[21:46:50 CEST] <cone-844> ffmpeg 03Michael Niedermayer 07master:e6c710600e1d: avcodec/svq3: add initialization which was removed from h264 init
[22:04:47 CEST] <cone-844> ffmpeg 03Anton Khirnov 07master:e1f907711a91: h264: factor out common code from init() and init_thread_copy()
[22:04:48 CEST] <cone-844> ffmpeg 03Michael Niedermayer 07master:abae926ef80c: Merge commit 'e1f907711a91e5ce19402a1831cfbe8f709b67f7'
[22:18:35 CEST] <cone-844> ffmpeg 03Anton Khirnov 07master:65afa65e7393: h264: drop redundant initialization of the scaling matrices
[22:18:36 CEST] <cone-844> ffmpeg 03Michael Niedermayer 07master:80a76905eb5c: Merge commit '65afa65e7393e7745427e267d6c6ca814c7c8b45'
[22:18:37 CEST] <cone-844> ffmpeg 03Michael Niedermayer 07master:aa000ddab235: avcodec/svq3: add scaling matrix init that was removed from h264 init
[22:25:23 CEST] <cone-844> ffmpeg 03Anton Khirnov 07master:a939e5b2527d: h264: remove a commented out nonsense line
[22:25:24 CEST] <cone-844> ffmpeg 03Michael Niedermayer 07master:097efa475056: Merge commit 'a939e5b2527d0c4628815b1d3d8e29ee921227e8'
[22:33:58 CEST] <cone-844> ffmpeg 03Anton Khirnov 07master:bd737b5178f3: h264: reset the private data in init_thread_copy()
[22:33:59 CEST] <cone-844> ffmpeg 03Michael Niedermayer 07master:39c6834c12a5: Merge commit 'bd737b5178f361a9b592691848f29a7a79603a7e'
[22:45:16 CEST] <cone-844> ffmpeg 03Michael Niedermayer 07master:31048d48b041: avcodec/libx265: export chosen picture types
[22:45:17 CEST] <cone-844> ffmpeg 03Martin Storsjö 07master:5637ff01218a: libavcodec: Clarify the documentation of the internal codec capability flags
[22:45:18 CEST] <cone-844> ffmpeg 03Michael Niedermayer 07master:db4ac46fed9b: Merge commit '31048d48b0410b9cfb6023054d81e627d01be342'
[22:45:19 CEST] <cone-844> ffmpeg 03Michael Niedermayer 07master:be920e4820ef: Merge commit '5637ff01218a44fcc1fff9b4382a15a6470abcc9'
[23:08:12 CEST] <cone-844> ffmpeg 03James Almer 07master:43482bd1a55c: swr/resample: use av_clip functions
[23:41:26 CEST] <ramiro> michaelni, what's the deadline for the qualification tasks?
[23:44:23 CEST] <michaelni> the number of slots we request has to be set by april 13th, we then have a bit more time to decide who gets the slots which we actually get from google
[23:44:55 CEST] <michaelni> ramiro, so it would be nice if qualification tasks with all review iterations would be done before the 13th
[23:45:52 CEST] <himangi> what is the purpose of using av_unints?
[23:46:06 CEST] <ramiro> michaelni, hm, and what about the outreachy qualifications?
[23:48:47 CEST] <michaelni> himangi, av_uninit is to silence compiler warnings
[23:49:15 CEST] <michaelni> if thats not documented ... then someoe should document it
[23:50:48 CEST] <michaelni> ramiro, outreachy deadline for applicants for which general funding is requested is 13th, for fundig thats ffmpeg specific there is more time
[23:51:17 CEST] <ramiro> hm, will we have ffmpeg-specific funding this time?
[23:51:38 CEST] <michaelni> ramiro, ask reynaldo
[23:52:02 CEST] <himangi> michaelni : but I see compiler warings like "used uninitialized whenever 'if' condition is false" for exactly those variables 
[23:52:40 CEST] <ramiro> michaelni, thanks
[23:53:21 CEST] <michaelni> np
[23:53:29 CEST] <michaelni> himangi, which compiler is taht ?
[23:54:29 CEST] <himangi> the one that fate uses
[23:54:58 CEST] <michaelni> which fate client ? locally or remote (fate.ffmpeg.org) ?
[23:55:27 CEST] <himangi> (http://fate.ffmpeg.org/log.cgi?time=20150405031205&log=compile&slot=arm64-darwin-clang-apple-5.1) libavformat/utils.c, pos_min and pos_max
[23:58:22 CEST] <michaelni> himangi, seems the av_uninit macro has no clang support
[23:58:51 CEST] <michaelni> iam not sure clang supports disabling this warning per variable
[23:59:01 CEST] <michaelni> if it does av_uninit should be extended
[23:59:54 CEST] <himangi> michaelni: I'll try that and probably add the documentation too
[00:00:00 CEST] --- Mon Apr  6 2015


More information about the Ffmpeg-devel-irc mailing list