[Ffmpeg-devel-irc] ffmpeg-devel.log.20170306

burek burek021 at gmail.com
Tue Mar 7 03:05:02 EET 2017


[00:06:12 CET] <unholy_me> @durandal_1707 I have written a prototype, how do I test it now?
[00:48:50 CET] <philipl> BtbN: are you in a position to test mpeg2 transcode? I'm see a weird thing here where it appears to get wedged in the hardware.
[00:49:02 CET] <philipl> Video Engine Utilization goes to 100%. Can't kill ffmpeg binary.
[00:49:13 CET] <philipl> This worked before.
[01:48:52 CET] <philipl> BtbN: blocked on a futex inside cuEventSynchronize inside cuvidDecodePicture
[01:54:06 CET] <philipl> managed to kill the process but leaves the GPU at 100% VEU. Need to reboot
[02:37:22 CET] <philipl> Don't get it. Seems to depend on the set of parameters passed or something. I wonder if there's some weird cuda context thing going wrong.
[02:37:50 CET] <philipl> Also can't seem to encode with bframes with the current driver version. It's returning an error from EncodePicture.
[02:37:58 CET] <philipl> Interested as to your experience.
[02:38:28 CET] <philipl> Ok. So bframe error goes away if I don't use hwaccel.
[02:38:37 CET] <philipl> So I guess this is somehow related to the filter changes.
[02:38:58 CET] <philipl> wm4: Seems like there's more cuvid/nvenc black magic hiding here.
[02:43:36 CET] <philipl> In another case I get a "No free surfaces" error if I use hwaccel with bframes and no problem without hwaccel
[10:54:55 CET] <cone-032> ffmpeg 03wang-bin 07master:3267e1703b32: avcodec/videotoolbox: set kCVPixelBufferOpenGLESCompatibilityKey for iOS
[11:08:19 CET] <cone-032> ffmpeg 03wm4 07master:a755b725ec1d: avcodec: consider an error during decoder draining as EOF
[11:08:20 CET] <cone-032> ffmpeg 03wm4 07master:808ab2fd0ba2: ffmpeg: remove unnecessary hack for decoders which refuse to drain
[11:08:21 CET] <cone-032> ffmpeg 03wm4 07master:ef86488696a8: avformat/flvdec: remove meaningless warning
[11:12:26 CET] <BtbN> philipl, I don't think I have ever tried mpeg2 decoding with cuvid, no.
[11:12:32 CET] <BtbN> Will have a look once I get home though
[14:07:03 CET] <kierank> wow now the aetey guy brings his own supporters
[14:14:02 CET] <wm4> where?
[14:16:34 CET] <kierank> kalle blomster
[14:20:15 CET] <Compn> atomnuker : put an alias in resample so if someone does -af resample it works ? i guess wm4 already said it
[14:21:56 CET] <atomnuker> yeah, I submitted both a patch to rename it and a separate patch to keep af_aresample but provide af_resample
[14:22:08 CET] <wm4> kierank: I wouldn't call it supportive
[14:22:37 CET] <wm4> I find "aresample" silly, but someone else said it's normal naming, so whatever
[14:58:40 CET] <BBB> atomnuker: so we have several applicants for the vp9 tasks now, shall I just tell the rest to do assembly as qual tasks?
[14:58:50 CET] <BBB> atomnuker: Im out of other options ATM ...
[14:59:05 CET] <BBB> maybe one of them can do something with other decoders
[15:25:02 CET] <atomnuker> I dunno which decoders need asm
[15:26:47 CET] <jkqxz> Cinepak?
[15:27:02 CET] <atomnuker> how SIMDd is FFV1?
[15:27:54 CET] <wm4> jkqxz: careful there...
[16:05:17 CET] <BBB> wm4: hey come on that was funny ;)
[16:05:23 CET] <BBB> wm4: you gotta be able to smile :)
[16:05:42 CET] <wm4> you never know what people will take seriously
[16:08:20 CET] <BBB> atomnuker: I dont know about ffv1, thats michaelnis territory& jpeg2k might be interesting, but its quite complex and Id like qual tasks to be easier...
[16:08:43 CET] <BBB> michaelni: do you think ffv1 has easy stuff left to be simd'ified?
[16:09:02 CET] <atomnuker> BBB: dirac lacks any 10bit assembly for any wavelets
[16:09:14 CET] <atomnuker> Haar should be easy enough for a quali
[16:09:18 CET] <BBB> thats about the same level of compelxity as jpeg2k right?
[16:09:20 CET] <BBB> oh, haar
[16:09:23 CET] <BBB> yeah that might be easy
[16:09:24 CET] <BBB> hm...
[16:09:25 CET] <BBB> ok
[16:09:36 CET] <BBB> do you want to lead one of the applicants through that?
[16:09:48 CET] <BBB> (Im already dealing with a number of them ATM :) )
[16:11:24 CET] <atomnuker> yeah, ok, do they have an IRC handle?
[16:11:49 CET] <BBB> not any of the ones whose email I havent answered in the past few days
[16:12:03 CET] <BBB> how about you pick ruben and see if hes interested
[16:12:03 CET] <atomnuker> you too, huh?
[16:12:14 CET] <BBB> :-p
[16:12:26 CET] <atomnuker> ok, but first I need to sleep
[16:13:04 CET] <BBB> haha ok
[16:16:59 CET] <unholy_me> @durandal_1707 I have declared a variable result= w+xcos+ysin, this should be stored in the frame right?
[16:18:08 CET] <michaelni> BBB, On the encoder side, i think there are some SIMD possibilities in FFV1, on the decoder side its harder
[16:18:34 CET] <BBB> michaelni: ok tnx, probably not quite right target then
[16:19:43 CET] <durandal_1707> unholy_me: you store processed samples in frame
[16:20:20 CET] <unholy_me> Okay, i have written a prototype how do I test it?
[16:21:38 CET] <durandal_1707> unholy_me: do you have b format samples?
[16:21:47 CET] <unholy_me> nope
[16:21:54 CET] <unholy_me> where can I download it?
[16:22:24 CET] <durandal_1707> find them, search on internet for ambisonia
[16:24:35 CET] <wm4> lol the CoC is _only_ used to throw supposed rule violations on each others heads
[16:25:01 CET] <kierank> only ffmpeg could use a behaviour policy for more misbehaviour
[16:26:04 CET] <unholy_me> @duarandal_1707 anyone else working on the Ambisonic project?
[16:26:18 CET] <unholy_me> doing the qualification tasks?
[16:46:16 CET] <durandal_1707> unholy_me: only you asked this much, others are either not dioing or keeping silent
[16:47:24 CET] <unholy_me> @durandal_1707 Sorry sir, this is my first time :
[16:48:19 CET] <durandal_1707> unholy_me: np
[16:48:49 CET] <cone-032> ffmpeg 03Michael Niedermayer 07master:a9f9b7f5c70e: avcodec/movtextdec: Call mov_text_cleanup() on close
[16:48:50 CET] <cone-032> ffmpeg 03Michael Niedermayer 07master:835d9f299cf6: avcodec/x86/cavsdsp: Put MMX code under mmx check
[16:48:51 CET] <cone-032> ffmpeg 03Michael Niedermayer 07master:8aa4f3bb4f77: avcodec/proresdec2: Do not discard frames with bad slices
[17:46:22 CET] <jamrial> at what point do we put an end to this cinepak madness? the guy spent a month trying to lobby his crap code in instead of using that time to actually fix it
[17:46:49 CET] <jamrial> now he's playing policy lawyer and crying about mistreatment. this guy can't be reasoned with
[17:46:49 CET] <nevcairiel> the problem is that he thinks he fixed it
[17:48:06 CET] <jamrial> that's whay i mean, he can't be reasoned with. the thread has stopped being technical days ago, so if he insists on not fixing his code then he needs to be banned/filtered from the ml
[17:48:15 CET] <jamrial> at this point it's noise and very disrupting
[18:06:32 CET] <BBB> the danger is that (as michaelni almost accidently did last month) someone will think that us ignoring him means the patch has no further comments and can thus be applied
[18:06:54 CET] <BBB> that would be incredibly dangerous, given how hard it is to revert patches in our tree
[18:22:07 CET] <Compn> invite him to irc so you can say it to him in real time
[18:36:02 CET] <BBB> Compn: but I already told him in one of my earlier emails
[19:40:07 CET] <jkqxz> The first rule on the CoC should be "do not attempt to invoke the CoC against other people on the ML".  It never helps.
[19:45:43 CET] <unholy_me> @durandal_1707 I have written a little buggy code for a square filter, will you have a look at it?
[19:46:10 CET] <durandal_1707> yes
[19:46:31 CET] <durandal_1707> unholy_me: pm me
[19:46:38 CET] <unholy_me> e-mail?
[19:47:10 CET] <durandal_1707> unholy_me: no private message over irc
[19:47:26 CET] <unholy_me> okay doing
[19:50:59 CET] <adeel1> durandal_1707: As the original code was giving incorrect image dimensions, I copied it from XBM. But, after remove it (again), the issue is still open.
[19:54:09 CET] <durandal_1707> adeel1: show what you have done?
[19:57:54 CET] <adeel1> durandal_1707: https://github.com/adl1995/FFmpeg/commit/9443794ffa6f101b56db2a65e9989e03adb9d90e
[20:05:54 CET] <durandal_1707> adeel1: notmal diff,not latest change
[20:13:35 CET] <adeel1> https://gist.github.com/adl1995/75e57f99886f6ba6a70315b62e43537c
[20:21:28 CET] <durandal_1707> adeel1: what error you get?
[20:33:00 CET] <adeel1> durandal_1707: I'm getting this: [xpm @ 0x7ff3dc007480] video_get_buffer: image parameters invalid [xpm @ 0x7ff3dc007480] get_buffer() failed
[20:34:26 CET] <durandal_1707>  
[20:34:59 CET] <durandal_1707> adeel1: get buffer must be called after is pix fmt set
[20:38:16 CET] <adeel1> durandal_1707: okay, the errors are now removed, but i'm getting:  18114 segmentation fault (core dumped
[20:39:12 CET] <durandal_1707> adeel1: because you need to learn pointers and arrays in C
[20:40:42 CET] <adeel1> durandal_1707: oh, let me check. I've actually played around with pointer in my previous courses, but it's been some time. will fix and let you know.
[21:00:41 CET] <cone-701> ffmpeg 03Kagami Hiiragi 07master:734d760e2fb2: lavc/libvpxenc: add -row-mt option
[21:45:29 CET] <durandal_1707> how much text this guy wrote he could rewrite swscale from scratch
[22:12:20 CET] <BtbN> Woah, a MPC-HC Update
[22:12:27 CET] <BtbN> Thought that project was dead
[22:16:28 CET] <peloverde> Their github has been pretty active for a while lately
[22:17:01 CET] <BtbN> Was about time they updated the bundled lavfilters and stuff
[22:17:05 CET] <JEEB> DShow is just not as fun as it was before
[22:17:21 CET] <BtbN> It still outperforms VLC by a lot for me
[22:17:24 CET] <BtbN> So I prefer it
[22:18:03 CET] <JEEB> most of my use cases don't require the UI and I got tired of teaching people how to fix the drivers trying to insert their filters into the rendering path
[22:18:14 CET] <JEEB> so I just moved to mpv as it finally had windows as a first-class citizen
[22:21:09 CET] <jamrial> BtbN: one of their core dev seems to have retired, so they are lacking in manpower
[22:21:29 CET] <jamrial> that's why it took them a year and three months to get a release out
[22:23:05 CET] <JEEB> most of the core devs left or became busy IRL
[22:23:13 CET] <JEEB> I think kasper93 is the only one left semi-active
[22:31:28 CET] <nevcairiel> it can probably keep functioning fine by just updating its lav version once in a while
[22:44:24 CET] <peloverde> Good support dpi scaling is probably important for windows players these days (and in fact SVG UI is one of the new features)
[00:00:00 CET] --- Tue Mar  7 2017


More information about the Ffmpeg-devel-irc mailing list