[Ffmpeg-devel-irc] ffmpeg-devel.log.20150415

[burek]

[03:34:48 CEST] <cone-487> ffmpeg 03Michael Niedermayer 07master:fbdaebb29861: avcodec/ituh263dec: Use check_marker()
[03:58:52 CEST] <jamrial> atomnuker: in case you didn't see it https://trac.ffmpeg.org/ticket/2686#comment:374
[04:01:46 CEST] <atomnuker> Once the patch makes it through to git master I'll respond, thanks
[04:57:20 CEST] <Prelude2004c> hey guys.. tried on ffmpeg and nobody answers so  i am going to try here.. using NVENC with 2 cards... only one gets used
[04:57:22 CEST] <Prelude2004c> how come ?
[04:57:37 CEST] <Prelude2004c> is there some way to say load balance NVENC through both GPUs or there is some param i can use to force it ?
[05:14:42 CEST] <Prelude2004c> can anyone help?
[05:31:33 CEST] <Prelude2004c> if nobody has the answer for NVENC .. does anyone know how to set the default GPU in linux ? in shell
[05:31:47 CEST] <Prelude2004c> if i maybe set the GPU prior to starting the ffmpeg session, maybe it will start in the correct one and i can manage that
[11:19:27 CEST] <wm4> kierank: bingo!
[11:19:44 CEST] <wm4> "If there is a point of the API that can not be accessed with the command-line tools, fix that first."
[11:26:30 CEST] <rcombs>  <Timothy_Gu> lglinskih: FFmpeg developers have lives :)
[11:26:34 CEST] <rcombs> ^ > implying
[11:56:51 CEST] <cone-716> ffmpeg 03Luca Barbato 07master:0cbb1473179d: x264: Map color parameters
[11:56:52 CEST] <cone-716> ffmpeg 03Michael Niedermayer 07master:0fbe18b63eab: Merge commit '0cbb1473179d35d4eb9c9446c8ed84e197e5389b'
[12:06:23 CEST] <cone-716> ffmpeg 03Luca Barbato 07master:ec89f35e3e5e: x265: Map color parameters
[12:06:24 CEST] <cone-716> ffmpeg 03Michael Niedermayer 07master:0842df2619a0: Merge commit 'ec89f35e3e5eb9ec5221266e801efeee7ecab4df'
[13:32:43 CEST] <cone-716> ffmpeg 03Michael Niedermayer 07master:a742a0536dbb: avformat/mp3dec: Allow forcing the use of the xing TOC for CBR files
[13:32:44 CEST] <cone-716> ffmpeg 03wm4 07master:ddd429fab635: Revert "avformat/mp3dec: offset seek index to end of id3v2 tag"
[13:32:45 CEST] <cone-716> ffmpeg 03wm4 07master:3163083f7b8f: avformat/mp3dec: offset seek index to end of vbr headers
[14:03:41 CEST] <kierank> wm4: wow
[14:03:55 CEST] <kierank> lglinskih: ignore wm4 and nicolas
[14:04:07 CEST] <kierank> well ignore nicholas
[14:04:12 CEST] <lglinskih> =)))
[14:05:03 CEST] <lglinskih> kierank: I wanted comments to my patch -- now I get them=)
[14:05:15 CEST] <kierank> lglinskih: personally I would not use the test source that they recommend
[14:05:28 CEST] <kierank> I would generate a simple square wave 
[14:05:34 CEST] <kierank> I want the API tests to be simple
[14:06:32 CEST] <kierank> lglinskih: but you'll need to add framecrc output so that it can be used in fate
[14:15:04 CEST] <wm4> kierank: shouldn't it use something other than framecrc?
[14:15:25 CEST] <kierank> well if it wants to go into fate, doesn't it have to?
[14:15:39 CEST] <wm4> you mean the framecrcenc pseudo-muxer, right? looks like a rather big potential error source
[14:15:48 CEST] <wm4> maybe run crc on the packet data manually
[14:16:01 CEST] <wm4> (it could be a shared utility function for all the tests)
[14:16:11 CEST] <kierank> yeah seems ok
[14:25:28 CEST] <Daemon404> kierank, for audio?
[14:25:38 CEST] <kierank> ?
[14:25:44 CEST] <Daemon404> re: fra,ecrc
[14:25:52 CEST] <kierank> for flac it's ok
[14:25:58 CEST] <kierank> which is the qualification task
[14:26:04 CEST] <Daemon404> ah
[14:27:43 CEST] <wm4> how many decoders can be made bitexact anyway?
[14:28:47 CEST] <Daemon404> wm4, id assume any except the audio codecs with no fixed point impl
[14:28:56 CEST] <Daemon404> unless there is some insane float codec
[14:28:58 CEST] <ubitux> wm4: speaking of seeking test, i wrote a "seek fuzzer" a while ago 
[14:29:13 CEST] <wm4> ubitux: sounds useful
[14:29:19 CEST] <ubitux> i'll dig it out if you want when i'm back
[14:29:29 CEST] <Daemon404> does anyone run afl btw?
[14:29:52 CEST] <ubitux> that would be nice
[14:29:55 CEST] <wm4> ubitux: lglinskih should be interested in it (at least if the gsoc application or whatever it was succeeds)
[14:31:29 CEST] <ubitux> wm4: yeah but i'm really short in time, so in one week or so, maybe
[14:33:51 CEST] <ubitux> and yeah afl looks nice; i think we could follow the logic of what that guy did with the sql thing
[14:34:17 CEST] <ubitux> use afl on every fate test basically
[14:34:21 CEST] <ubitux> that could be a nice project
[14:34:29 CEST] <ubitux> anyway, going away
[14:34:36 CEST] <wm4> what's afl?
[14:43:21 CEST] <kierank> a fuzzer
[14:44:44 CEST] <Daemon404> american fuzzy lop
[14:44:47 CEST] <Daemon404> stupid name
[14:44:49 CEST] <Daemon404> good fuzzer
[14:45:46 CEST] <wm4> sounds pretty interesting
[15:01:59 CEST] <lglinskih> kierank: ok, I don't understand exactly how to do it (I don't know much about fate), but I'll read and ask you later)
[15:14:15 CEST] <kierank> fate basically is a script that compares the crc of the decoded file to a reference
[15:14:24 CEST] <kierank> you should have a look at the fate instructions on ffmpeg.org
[15:17:53 CEST] <wm4> michaelni: I might try to add a test for gapless - can the sample file I posted be added to fate, or is it too big?
[16:45:42 CEST] <michaelni> wm4, 250kb AG.mp3 ? that would be ok, in which directory should i put it ?
[16:46:45 CEST] <wm4> ok... no need to do add it yet, I first have to write a patch
[17:56:41 CEST] <wm4> is there a test yet which includes AVFrame or AVPacket side data?
[17:57:29 CEST] <michaelni> i think so but i dont remember which
[18:00:03 CEST] <Prelude2004c> good day everyone. Tried yesterday but got no responses. Can someone point me in the right direction please. I am using NVENC to encode and i ahve 2 cards in my system. I don't know the flag to have the system load balance between both cards as right now it is only using 1 GPU. Can anyone help?
[18:00:31 CEST] <wm4> Prelude2004c: enter a support contract with nvidia
[18:00:53 CEST] <nevcairiel> the nvenc encoder has a "gpu" option which lets you select which gpu is being used
[18:00:58 CEST] <nevcairiel> 0, 1, 2 etc
[18:01:02 CEST] <nevcairiel> thats all we can do
[18:01:06 CEST] <nevcairiel> otherwise, ask nvidia
[18:01:18 CEST] <Prelude2004c> really? its some param ? i tried looking everywhere on google. nobody posts it as an example
[18:01:31 CEST] <Prelude2004c> i hope its something simple like a param that i set
[18:01:35 CEST] <nevcairiel> you should really read the help output from ffmpeg
[18:01:35 CEST] <Prelude2004c> but not sure where to set it
[18:02:58 CEST] <Prelude2004c> i looked at all the ffmpeg doc, nvidia nvenc stuff.. nothing does it say to select GPU do this.
[18:03:08 CEST] <Timothy_Gu> Prelude2004c: yes it's -gpu
[18:03:34 CEST] <kepstin-laptop> Prelude2004c: you set it in the same way as any other encoder avoption.
[18:04:22 CEST] <Prelude2004c> really?
[18:04:26 CEST] <Prelude2004c> hum.. i will try thank you
[18:04:54 CEST] <Timothy_Gu> Prelude2004c: by default it uses the gpu with the most CUDA computing unit
[18:06:25 CEST] <Prelude2004c> you are the best
[18:06:27 CEST] <Prelude2004c> thank you
[18:06:36 CEST] <Prelude2004c> ya but cards are identical
[18:06:45 CEST] <Prelude2004c> one has like 90% load and the other is 0
[18:06:51 CEST] <Prelude2004c> setting up -GPU worked
[18:06:52 CEST] <Prelude2004c> :)
[18:06:56 CEST] <Prelude2004c> so much thank you's
[18:07:09 CEST] <Prelude2004c> i was at it for a few hours .. couldn't find anything on it
[18:19:37 CEST] <philipl> You cannot split an encoding job across GPUs. That's not how the hardware works. What you're attempting to do cannot be done.
[18:19:46 CEST] <philipl> It's not anything like multi-threaded CPU encoding.
[18:20:03 CEST] <philipl> The hardware has decided silicon for encoding with a fixed pipeline.
[18:39:02 CEST] <Timothy_Gu> Prelude2004c: yeah ffmpeg doesn't have load balancing..
[18:39:30 CEST] <Timothy_Gu> philipl: no he's trying to have multiple ffmpeg instances using different cards
[18:46:48 CEST] <philipl> Timothy_Gu: ah.
[19:23:49 CEST] <jc0n> hey if #ffmpeg-user is empty and I pop in here to ask a quick question about ffserver are you guys gonna kill me?
[19:25:32 CEST] <lordkrondor> #ffmpeg seems like the place to ask questions
[19:25:39 CEST] <jc0n> alright thank you
[19:56:50 CEST] <cone-716> ffmpeg 03Michael Niedermayer 07master:574978d2bdbc: avcodec/intelh263dec: Use check_marker()
[19:56:51 CEST] <cone-716> ffmpeg 03Nedeljko Babic 07master:f4ccf3890285: libavutil/softfloat: Change order of fields in SoftFloat structure.
[20:15:46 CEST] <wm4> so why does "ffmpeg -ss 0 -i some.mp3" not seek to 0, but ts=353604?
[20:16:14 CEST] <wm4> it seems to include the start time (who knows why the heck gapless was designed to move the start time from 0 to a later timestamp)
[20:16:59 CEST] <wm4> and it seems to do so unconditionally
[20:25:56 CEST] <wm4> is this acceptable for a fate test? http://sprunge.us/IOSI
[21:03:57 CEST] <cone-716> ffmpeg 03Rostislav Pehlivanov 07master:c5d4f87e8111: aaccoder: Implement Perceptual Noise Substitution for AAC
[00:00:00 CEST] --- Thu Apr 16 2015