[Ffmpeg-devel-irc] ffmpeg-devel.log.20150929

burek burek021 at gmail.com
Wed Sep 30 02:05:02 CEST 2015


[00:33:23 CEST] <atomnuker> durandal_1707: yeah, I am
[00:34:14 CEST] <atomnuker> laptop broke down just as soon as I got it working so I haven't been able to make comparisons yet
[04:45:58 CEST] <cone-198> ffmpeg 03Thierry Foucu 07master:7f72f2d75e58: libavformat/flvdec.c: free always the packet after a resync.
[05:02:46 CEST] <cone-198> ffmpeg 03Ganesh Ajjanagadde 07master:b3066be0e460: avcodec/videotoolbox: fix -Wunused-but-set-variable
[09:12:08 CEST] <saste_> I dislike the way the GSoC notice is shown on the website
[09:17:28 CEST] <fritsch> saste_: impressive tasks
[11:36:27 CEST] <saste_> really, are we going to translate the wiki to chinese? ^^
[11:46:45 CEST] <wm4> PHP? wtf
[12:18:05 CEST] <blez> hello
[12:18:19 CEST] <blez> I get "Unknown Error" using ffmpeg, how can I see what's the actual error?
[12:36:24 CEST] <atomnuker> Who wrote this stuff? It's wrong
[12:36:41 CEST] <wm4> what stuff
[12:36:43 CEST] <wm4> ffmpeg?
[12:36:50 CEST] <atomnuker> TNS worked pretty good for low bitrates up until Claudio's RC push broke it
[12:37:09 CEST] <atomnuker> (nobody say anything about that, it'll probably delay him another week)
[12:37:51 CEST] <atomnuker> whatever, I can't be bothered to care
[12:52:50 CEST] <iive> atomnuker: i think you should tell him.
[13:05:39 CEST] <durandal_170> atomnuker: can you fix it?
[13:06:39 CEST] <durandal_170> btw, you have working machine?
[13:28:13 CEST] <superware> I'm playing a live h264 over udp stream, but sometimes (network glitches etc) the stream hangs indefinitely. Restarting ffmpeg continues playback. Since it's very hard to reproduce, I've complied a very messed up stream which reproduces the issue, please see log http://pastebin.com/QuA5nCvQ up to line 229 where is some video, but after that - nothing.
[13:41:25 CEST] <cone-619> ffmpeg 03Michael Niedermayer 07master:1b82b934a166: avcodec/x86/sbrdsp: Fix using uninitialized upper 32bit of noise
[13:46:02 CEST] <cone-619> ffmpeg 03Rémi Denis-Courmont 07master:b10b6ac7a902: vdpau: deprecate av_vdpau_get_profile()
[13:46:03 CEST] <cone-619> ffmpeg 03wm4 07master:a5d58fea68b9: lavc: reimplement avcodec_get_type() using codec descriptors
[13:46:04 CEST] <cone-619> ffmpeg 03wm4 07master:cc8db760616a: mpegts: use avcodec_get_type() to set codec_type
[13:46:05 CEST] <cone-619> ffmpeg 03wm4 07master:a41e5e192ed8: vdpau: fix constrained baseline fallback
[13:46:06 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:b123d82c4182: Merge commit 'b10b6ac7a902f28e09e37a29c392e2f0c19e9526'
[13:46:07 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:2f63e83c26f1: Merge commit 'a5d58fea68b9212e0065a71939e921505504a9bb'
[13:46:08 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:4b7685b6c626: Merge commit 'cc8db760616a7ec3bd39b22ca45888c01326db13'
[13:46:09 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:d0bf83ff1dd3: Merge commit 'a41e5e192ed8f79f6607f978dee3205580ba5039'
[13:46:34 CEST] <superware> anyone? :|
[13:48:07 CEST] <kierank> use latest ffmpeg
[13:48:21 CEST] <cone-619> ffmpeg 03Henrik Gramner 07master:5405584b7b54: checkasm: Use a self-balancing tree
[13:48:22 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:fd0fa9af80fe: Merge commit '5405584b7b54ca889c341743de1d58792449830d'
[13:48:30 CEST] <kierank> i.e ffmpeg git
[13:48:38 CEST] <cone-619> ffmpeg 03Henrik Gramner 07master:cc2855210000: checkasm/x86: Correctly handle variadic functions
[13:48:39 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:88dbdf4845e3: Merge commit 'cc285521000020ab237d183dc3a26f99fce3030f'
[13:51:19 CEST] <superware> kierank: thanks, tried that, same...
[13:51:36 CEST] <kierank> open it in gdb and see where it blocks
[13:52:00 CEST] <kierank> I used to get these issues a lot
[13:52:35 CEST] <nevcairiel> i think his definition of hang is different, ffmpeg seems to still function, it just doesnt output frames anymore
[13:52:47 CEST] <superware> I don't think the code blocks, no video frame is playing after "the event", you can see the log keeps going...
[13:52:52 CEST] <kierank> ah
[13:52:58 CEST] <kierank> different bug to mine perhaps
[13:53:09 CEST] <superware> it's as if a specific packet/frame causes FFmpeg to enter a certain state from which it cannot recover
[13:53:13 CEST] <cone-619> ffmpeg 03Anton Khirnov 07master:ae25413daf42: lavfi: do not exclude hwaccel formats from ff_all_formats()
[13:53:14 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:c36e85b3d9a1: Merge commit 'ae25413daf42a06f078ed81bb545ec23a8e0b482'
[13:54:41 CEST] <superware> the relevant lines from the log: reference count 2 overflow; reference picture missing during reorder; reference count overflow; decode_slice_header error; mmco: unref short failure; number of reference frames (0+2) exceeds max (1; probably corr
[13:54:41 CEST] <superware> upt input), discarding one; ...
[13:54:48 CEST] <kierank> that's just packet loss errors
[13:55:18 CEST] <superware> but it only happens when the issue occurs
[13:56:20 CEST] <kierank> anyway run in gdb and get a trace for all the threads where it hangs imo
[13:56:21 CEST] <superware> before that the video is very choppy (to non-existent) but none of those are reported
[13:56:25 CEST] <kierank> and open a ticket
[13:57:42 CEST] <superware> I don't think I have sufficient knowledge to debug ffmpeg h264
[13:57:46 CEST] <DHE> superdump: does 4 minutes, 20 seconds sound like the time before the stream shuts down? or is it more erratic?
[13:58:28 CEST] <DHE> (this assumes you're using multicast)
[13:59:09 CEST] <superware> superdump is here, but not me :)
[13:59:34 CEST] <superware> it happens after 1 minute and 38 seconds.
[14:01:41 CEST] <superware> I couldn't reproduce in normal operation, so I recorded the network pcap and discarded half of the packets.
[14:02:26 CEST] <DHE> oops
[14:03:24 CEST] <superware> now I have a file which I'm playing via PlayCap, ffplay starts playing some very messed up frames, but always stops outputting frames at the same time. restarting ffplay continues with the bad video.
[14:04:21 CEST] <DHE> so, half of 4:20....
[14:04:31 CEST] <superware> naa :)
[14:04:42 CEST] <DHE> ah, not quite...
[14:04:52 CEST] <DHE> sorry it's still early
[14:06:06 CEST] <superware> do you know of someone that has h264 experience?
[14:06:31 CEST] <DHE> me?
[14:07:10 CEST] <durandal_170> superware: could you upload file you give to ffplay?
[14:07:25 CEST] <cone-619> ffmpeg 03Anton Khirnov 07master:741b352b16da: qsvdec: list supported pixel formats
[14:07:26 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:4554178f072e: Merge commit '741b352b16dad74b87c4a39bade8902633a2b0e6'
[14:07:56 CEST] <durandal_170> does it happens with -threads 1?
[14:08:07 CEST] <DHE> it sounds to me like you've got a multicast stream that's possible encrypted, and missing an IGMP querier
[14:08:37 CEST] <superware> durandal_170: uploading to expirebox.com
[14:09:30 CEST] <superware> plus checking -threads 1
[14:10:36 CEST] <superware> http://expirebox.com/download/55e58de3de92c1203dc8a220ea7b1b76.html
[14:10:55 CEST] <cone-619> ffmpeg 03Anton Khirnov 07master:83847cc8fa97: qsvenc: do not try to close the encoder if the session is NULL
[14:10:56 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:49a25d02ddf6: Merge commit '83847cc8fa97e0fc637a0962bafb837acdb6eacc'
[14:12:07 CEST] <superware> durandal_170: -threads 1 didn't help
[14:14:04 CEST] <cone-619> ffmpeg 03Anton Khirnov 07master:5d2daebf3cc8: qsvenc: mark the encoders as INIT_CLEANUP
[14:14:05 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:66682145216f: Merge commit '5d2daebf3cc8de4cee1973db6a2229beaad3b7cd'
[14:14:31 CEST] <cone-619> ffmpeg 03Anton Khirnov 07master:d0c8c380ecf3: qsv: document AVQSVContext members
[14:14:32 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:029aa8ff1407: Merge commit 'd0c8c380ecf3d9bb16621a4fb59ebbcde301002a'
[14:14:40 CEST] <superware> durandal_170: could you download it?
[14:17:21 CEST] <cone-619> ffmpeg 03Anton Khirnov 07master:2c32eace5ec4: qsvdec: close the MFX decoder on uninit
[14:17:22 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:9457a11a220c: Merge commit '2c32eace5ec4d1d7ca4e0220856cd2815ccc71b2'
[14:17:46 CEST] <cone-619> ffmpeg 03Anton Khirnov 07master:3ee462dca103: examples/qsvdec: do not free the surfaces in the frame_free() callback
[14:17:47 CEST] <cone-619> ffmpeg 03Anton Khirnov 07master:539e5ac2ecfe: examples/qsvdec: free the lavc decoder before closing MFX/VAAPI
[14:17:48 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:d0a1605134a8: Merge commit '3ee462dca1038e63b8e8d5e751121736d5772a5d'
[14:17:49 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:b1284a4b4a6c: Merge commit '539e5ac2ecfec2e2f441222a43fb0583643ea607'
[14:19:03 CEST] <durandal_170> superware: downloading it, how I can use it?
[14:19:40 CEST] <superware> PlayCap
[14:20:39 CEST] <cone-619> ffmpeg 03Anton Khirnov 07master:8aecec84021a: qsvdec: make ff_qsv_decode_init() static
[14:20:40 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:27673f1bea1a: Merge commit '8aecec84021a61b943718ff3d7c2c57fcd4af199'
[14:24:24 CEST] <cone-619> ffmpeg 03Gregory J. Wolfe 07master:1a4c5fe56008: libopenh264enc: Use av_log() to log messages
[14:24:25 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:4d9b78a875b2: Merge commit '1a4c5fe56008c61b0362c75bea5d446dc5b256bc'
[14:26:59 CEST] <durandal_170> superware: could you open bug report on tracker?
[14:27:41 CEST] <cone-619> ffmpeg 03Luca Barbato 07master:678f788fea33: configure: Set the initial ldflags to match the cflags
[14:27:42 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:071f6edb985d: Merge commit '678f788fea3380e5cbbf75baac5cc0ce07a56a42'
[14:35:45 CEST] <cone-619> ffmpeg 03Luca Barbato 07master:1016a75cf317: configure: mips: Support mips r6, r2 and r1
[14:35:46 CEST] <cone-619> ffmpeg 03Vicente Olivert Riera 07master:d00bb8addccb: mips: intreadwrite: Only execute that code for mips r1 or r2
[14:35:48 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:c096226ae614: Merge commit '1016a75cf3170648dc9b59fdef170cbfc142f8ad'
[14:35:49 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:a2426798d602: Merge commit 'd00bb8addccb63fa3feacb06d2a310731dc0113b'
[14:35:56 CEST] <saste_> who's the website maintainer?
[14:36:50 CEST] <saste_> there is this annoying margin-top: 180px attribute for h4 elements, how can I fix it?
[14:37:33 CEST] <wm4> there is a name I've never seen before in MAINTAINERS for the website
[14:38:31 CEST] <nevcairiel> isnt ubitux to blame for the recent redesign
[14:38:35 CEST] <nevcairiel> or do i mis-remember
[14:39:13 CEST] <ubitux> BBB: i'm sorry i won't have time
[14:39:17 CEST] <ubitux> wm4: api, not abi (subtitles)
[14:39:20 CEST] <BBB> oh youre still alive!
[14:39:23 CEST] <BBB> no problem
[14:39:24 CEST] <ubitux> nevcairiel: not really
[14:39:24 CEST] <Compn> the html of the website is in ffmpeg-web repo
[14:39:31 CEST] <ubitux> nevcairiel: db0 is the author
[14:39:41 CEST] <Compn> saste_ : i can commit to ffmpeg-web repo if you have a patch or just tell me what to edit
[14:40:12 CEST] <ubitux> grep website MAINTAINERS
[14:40:16 CEST] <saste_> Compn, I know where the ffmpeg web repository is
[14:40:32 CEST] <saste_> I wanted to ask, is she on IRC?
[14:41:12 CEST] <ubitux> she was, seems she hasn't been for a while
[14:41:20 CEST] <superware> durandal_170: sure, but I first want to verify it's a real issue, can you please try to reproduce?
[14:41:53 CEST] <saste_> there is something about the h4 attribute which is not immediately obvious, I can't grep for top-margin on the local repo
[14:43:51 CEST] <ubitux> BBB: yeah, surviving :)
[14:44:05 CEST] <BBB> awh :(
[14:44:25 CEST] <ubitux> (joking, i'm fine, just heavy load on various aspects)
[14:48:21 CEST] <superware> durandal_170?
[14:51:53 CEST] <superware> are there several different h264 decoders in ffmpeg?
[14:53:22 CEST] <wm4> there's only one, although there are several other wrappers for hw decoding APIs which decode h264
[14:54:20 CEST] <durandal_170> superware: I can't apt-get deps for playcap
[14:55:29 CEST] <superware> durandal_170: are you on Windows?
[15:08:45 CEST] <BBB> does anyone remember if theres an instruction on x86 (sse2 or so) to move just the higher half of a xmm register from one reg to another, but leave the lower half untouched?
[15:09:21 CEST] <BBB> (or move lower half and leave higher bits untouched)
[15:10:04 CEST] <BBB> the best I can think of is shufps but Im wondering if theres something better
[15:10:09 CEST] <nevcairiel> there is movhlps which moves the high-quadword from origin to the low-quadword of dest, without touching dests high-quadword, but i guess thats not quite what you need
[15:11:22 CEST] <nevcairiel> (and the reverse also exists)
[15:11:50 CEST] <BBB> right, thats low to high, I need high to high or low to low
[15:12:02 CEST] <BBB> shuffling is such a pain on x86 sometimes
[15:12:11 CEST] <BBB> too many instructions that do just a small subset of the general thing
[15:12:12 CEST] <nevcairiel> movhps can load the high quad from memory
[15:12:28 CEST] <nevcairiel> still no good eh?
[15:12:39 CEST] <BBB> hehehe right
[15:12:42 CEST] <BBB> I want movhps from reg to reg
[15:13:01 CEST] <durandal_170> superware: no, ubuntu
[15:13:05 CEST] <BBB> punpcklqdq is sort of like movlhps, in fact
[15:13:25 CEST] <BBB> but still not what I want
[15:16:01 CEST] <cone-619> ffmpeg 03Ganesh Ajjanagadde 07master:d3e5fbb14069: avcodec/apedec: fix undefined left shifts of negative numbers
[15:21:28 CEST] <nevcairiel> i think its funny if warning fixes introduce new warnings
[15:21:35 CEST] <nevcairiel> someone wasnt very good at fixing
[15:21:54 CEST] <wm4> trollol
[15:23:19 CEST] <cone-619> ffmpeg 03wm4 07master:948f3c19a8bd: lavc: Make AVPacket.duration int64, and deprecate convergence_duration
[15:23:20 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:b01891a9f08b: Merge commit '948f3c19a8bd069768ca411212aaf8c1ed96b10d'
[15:23:20 CEST] <superware> durandal_170: is this ok? http://trac.ffmpeg.org/ticket/4893
[15:23:25 CEST] <Daemon404> every valid line of C code produces a warning with some C conpiler
[15:23:37 CEST] <nevcairiel> gcc is like our main target
[15:23:42 CEST] <martijnb> I thought the challenge was to introduce increasingly obscure warnings until no one can understand it anymore
[15:24:02 CEST] <Daemon404> nevcairiel, holds true hetween gcc versions
[15:24:09 CEST] <Daemon404> between*
[15:26:04 CEST] <BBB> martijnb: mission accomplished!
[15:27:35 CEST] <cone-619> ffmpeg 03Ganesh Ajjanagadde 07master:977f41e274a6: mlpdec: Fix a undefined left shift of negative number
[15:27:36 CEST] <cone-619> ffmpeg 03Ganesh Ajjanagadde 07master:4885bde3187a: motion_est_template: Fix undefined left shift of negative number
[15:27:37 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:cbed8dbb7e1e: Merge commit '977f41e274a66c9d257186ca1df8373a09cc4d40'
[15:27:38 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:fe294b34030b: Merge commit '4885bde3187a2bb0cae85b67796e07db233bf77f'
[15:29:20 CEST] <cone-619> ffmpeg 03Ganesh Ajjanagadde 07master:84dfc426cea7: avresample: Remove an unused variable
[15:29:21 CEST] <cone-619> ffmpeg 03Vittorio Giovara 07master:cf7d2f2d2134: lavc: Simplify checking quant bias option
[15:29:23 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:24fea8998375: Merge commit '84dfc426cea7242099aea9d47121cea65dffd936'
[15:29:24 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:773570a9dce9: Merge commit 'cf7d2f2d2134c0854edf2db91e7436ac2bc9874f'
[15:30:34 CEST] <cone-619> ffmpeg 03Vittorio Giovara 07master:3973f0f773e0: Revert "avconv_opt: Allow printing private options"
[15:30:35 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:e490fee6b971: Merge commit '3973f0f773e0bd212734eccda78aa798f8b20692'
[15:30:55 CEST] <BBB> nevcairiel: hey since you know x86 asm, can you review my loopfilter 10/12bpp vp9 patch?
[15:31:02 CEST] <BBB> nevcairiel: much more exciting than hevc
[15:31:06 CEST] <cone-619> ffmpeg 03Vittorio Giovara 07master:b5e4f393b675: avconv: Make the private options discovery more manifest
[15:31:07 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:7255cf003ef1: Merge commit 'b5e4f393b6757629281f58c3f3f6d55ca522ab60'
[15:31:19 CEST] <nevcairiel> i know basic asm/simd, but my brain always starts to hurt when i try to understand complex algorithms
[15:31:41 CEST] <BBB> loopfilter is very simple
[15:31:51 CEST] <BBB> you just close your eyes and imagine pink flying ponies
[15:31:54 CEST] <BBB> thats a loopfilter
[15:36:25 CEST] <cone-619> ffmpeg 03Thierry Foucu 07master:c5e5e6306223: riff: Add support for RV40 codec in AVI
[15:36:26 CEST] <cone-619> ffmpeg 03Christophe Gisquet 07master:c49cbecbae5a: dnxhddec: Decode and use interlace mb flag
[15:36:27 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:b3bf41cb75d4: Merge commit 'c5e5e6306223623de8352a3ecd224956aa5beb37'
[15:36:28 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:1342d7b2f941: Merge commit 'c49cbecbae5a42f4ca004197b0118cc50aaaca2e'
[15:37:41 CEST] <ubitux> ffmpeg -ss 30 -f lavfi -i testsrc -ss 20 -t 10 -y out.avi     50 ’ 60
[15:37:43 CEST] <ubitux> ffmpeg -ss 30 -f lavfi -i testsrc -ss 20 -to 60 -y out.avi    50 ’ 90
[15:37:46 CEST] <ubitux> is this a wanted behaviour?
[15:38:34 CEST] <cone-619> ffmpeg 03Vittorio Giovara 07master:e94e651c762f: dnxhddec: Enable frame threading
[15:38:35 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:b03260fde08c: Merge commit 'e94e651c762f90ac5fd2dc9bd3ba1336a77d5b5c'
[15:44:27 CEST] <cone-619> ffmpeg 03Vittorio Giovara 07master:c9943f00cfa2: vf_framepack: Use av_image_copy() where appropriate
[15:44:28 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:f35c4ede9e0e: Merge commit 'c9943f00cfa2471d1b8a3a9ddc7a21049a71090e'
[15:45:09 CEST] <cone-619> ffmpeg 03Vittorio Giovara 07master:26e8fa3b508e: tiny_psnr: Use the correct abs() version
[15:45:10 CEST] <cone-619> ffmpeg 03Hendrik Leppkes 07master:fc97b1f091ff: Merge commit '26e8fa3b508eb047e85f4e923fc8e82a1aa656ba'
[15:45:32 CEST] <nevcairiel> BBB: i'll give it a look, but i cant promise much
[15:46:12 CEST] <BBB> nevcairiel: well, more generally, Im looking for someone tor review anything assembly at all
[15:46:22 CEST] <BBB> Ill have a few more coming up and nobodys reviewing, which kind of sucks
[15:47:05 CEST] <durandal_170> I just commit what nobody reviews
[15:51:58 CEST] <saste_> ubitux, are you optimizing testsrc or what?
[15:52:31 CEST] <ubitux> saste_: no, just wondering about the inconsistency between -t and -to
[15:52:42 CEST] <superware> is cehoyos here?
[15:52:50 CEST] <Daemon404> not usually
[15:53:08 CEST] <superware> ok
[16:29:50 CEST] <nevcairiel> getting rid of the convergence_duration hackery from my own code sure made my feel good
[16:29:56 CEST] <nevcairiel> s/my/me/
[16:33:45 CEST] <wm4> oh it's already in?
[16:34:13 CEST] <nevcairiel> an hour ago or so
[16:47:32 CEST] <BBB> I never understood why convergence_duration was introduced
[16:47:50 CEST] <nevcairiel> i think it was introduced for something else entirely
[16:48:00 CEST] <nevcairiel> it was then abused as a 64-bit duration field for subtitles
[16:48:03 CEST] <BBB> it seemed like somebody mistook a doxy comment for abi
[16:48:23 CEST] <BBB> hm, ok
[16:48:34 CEST] <nevcairiel> the comment still states its original concept
[16:48:40 CEST] <nevcairiel> or well used to until it was removed
[16:48:54 CEST] <Paranoialmaniac> i though it is used for gradual decoder refresh but...
[16:49:06 CEST] <nevcairiel> thats what it kinda was for
[16:49:18 CEST] <nevcairiel> time until the stream "converges", ie. when the refresh is done
[16:49:25 CEST] <nevcairiel> but it was never used for that
[17:24:37 CEST] <Gramner> BBB: for merging parts of registers; shufps with SSE, pblendw/blendps with SSE4.1, vpblendd with AVX2
[17:24:49 CEST] <BBB> ah, blend
[17:24:54 CEST] <BBB> ok, this is sse2 targetted, so shufps it is
[17:24:56 CEST] <BBB> ty
[17:25:20 CEST] <Gramner> in AVX-512 you can use an opmask blend with every instruction as well
[17:25:52 CEST] <durandal_1707> BBB: I want to write simd for maskedmerge filter
[17:26:54 CEST] <BBB> Gramner: uh ok that sounds a little overkill
[17:26:58 CEST] <BBB> typically intel I guess
[17:27:13 CEST] <BBB> durandal_1707: algo desc in C? and did you dspize the function?
[17:27:32 CEST] <BBB> (i.e. put it in a blaDSPContext in its own file with a nice header etc., and write a checkasm test, etc.)
[17:28:09 CEST] <durandal_1707> not yet
[17:28:56 CEST] <BBB> px1*weight+px2*(1-weight)?
[17:29:00 CEST] <BBB> (in fixed point)
[17:29:07 CEST] <Gramner> I'd say it's typical intel to invent this completely new backwards-incompatible future-proof instruction encoding called VEX (AVX), just to replace it again a few years later with EVEX. as a matter of fact, they had probably dropped VEX internally before hardware using it was even released to the general public
[18:41:43 CEST] <nevcairiel> so much for future proof =p
[18:53:34 CEST] <wm4> is anyone against applying that videotoolbox patch from Stefano Pigozzi?
[18:59:52 CEST] <TheRyuu> https://sourceware.org/bugzilla/show_bug.cgi?id=19011 <---do my eyes decieve me, sanity from gnu?
[19:01:46 CEST] <Gramner> TheRyuu: nice
[19:03:18 CEST] <JEEB> wow
[19:03:44 CEST] <TheRyuu> he's gonna be sad when it's a patch series though (if I do it)
[19:05:37 CEST] <cone-619> ffmpeg 03Michael Niedermayer 07master:ed18c49f5ff8: fate: Add basic license header check
[19:07:41 CEST] <jamrial> ^ i'd say tools/patcheck is the proper way to check this and not fate, but it wouldn't work for libav merges which is the source of the wrong headers, so guess it's ok
[19:10:41 CEST] <wm4> fate now tests source code too!
[19:11:36 CEST] <wm4> (I think someone wants to force someone else to change the project name on merges?)
[19:14:37 CEST] <durandal_1707> thats ok
[19:16:42 CEST] <durandal_1707> ubitux: is it ok to have assumefps?
[19:24:02 CEST] <DHE> there, hopefully that'll be the last revision of the a53 CC patch
[19:35:32 CEST] <cone-619> ffmpeg 03PrzemysBaw Sobala 07master:01dd7e025c24: lavf/img2dec: Fix memory leak
[19:49:38 CEST] <cone-619> ffmpeg 03Ganesh Ajjanagadde 07master:308e7484a3b1: avcodec/x86/rnd_template: silence -Wunused-function on --disable-mmx
[20:09:35 CEST] <ubitux> durandal_1707: what is it?
[20:13:02 CEST] <durandal_1707> changes fps and pts without dropping or adding frames
[20:25:55 CEST] <ubitux> durandal_1707: what would be the difference with setpts?
[20:26:39 CEST] <durandal_1707> setpts changes framerate?
[20:27:12 CEST] <ubitux> no because of the expression i guess
[20:28:01 CEST] <ubitux> what field are we talking about?
[20:28:07 CEST] <ubitux> we have a link->framerate or sth?
[20:29:38 CEST] <durandal_1707> frame_rate
[20:30:02 CEST] <ubitux> just add a setframerate filter maybe
[20:30:11 CEST] <ubitux> that just changes this field
[20:30:30 CEST] <ubitux> but well, then adding a setpts filter would be annoying
[20:30:43 CEST] <ubitux> yeah well, dunno, no opinion
[20:32:26 CEST] <durandal_1707> setfps is shorter
[20:48:00 CEST] <wm4> why are the old vdpau hwaccels (h264_vdpau etc.) still not gone?
[20:53:58 CEST] <durandal_1707> nobody removed them
[20:59:02 CEST] <cone-619> ffmpeg 03Paul B Mahol 07master:a01914924969: avfilter/vf_atadenoise: do not use uninitialized data
[21:13:22 CEST] <iive> somebody said that libav have delayed ff_api_vdpau once again, so we are doing that too.
[21:13:56 CEST] <nevcairiel> welcome to 4 weeks ago
[21:14:41 CEST] <nevcairiel> the deprecation just isnt that old yet
[21:17:13 CEST] <wm4> Libav have removed these codecs ages ago
[21:17:23 CEST] <wm4> what they didn't remove are some API symbols
[21:17:25 CEST] <durandal_1707> JEEB: tried zscale?
[21:17:38 CEST] <BBB> wm4: I believe the deprecation works functionally, so if you dont like that feature, just use -DFF_API_blabla=0 as extra-cflags=..; configure arg
[21:17:41 CEST] <wm4> (like CODEC_CAP_HWACCEL_VDPAU)
[21:18:11 CEST] <JEEB> durandal_1707: I did before, I should test a newer version
[21:18:11 CEST] <wm4> but hey lets keep this crap because we're so misinformed!
[21:19:04 CEST] <nevcairiel> noone intentionally prolonged its life, but noone send a patch to remove it either
[21:19:13 CEST] <nevcairiel> someone put it under the same deprecation guards as the  API
[21:19:16 CEST] <nevcairiel> so now its tied to it =p
[21:19:48 CEST] <wm4> how about a patch that removes the implementation
[21:19:58 CEST] <wm4> or is going through pointless untested ifdeffery mandatory
[21:22:16 CEST] <nevcairiel> its a decoder, the existance of a decoder is not API
[21:23:29 CEST] <wm4> so I'll send a patch to remove them tomorrow (without the API)
[21:23:38 CEST] <wm4> and then it'll get rejected and I'll be angry
[21:24:07 CEST] <durandal_1707> JEEB: its always much slower than swscale
[21:24:32 CEST] <durandal_1707> wm4: voting!
[21:24:51 CEST] <BtbN> Hm, should I just merge the OpenCL stuff I made? Nobody seems to care.
[21:25:23 CEST] <durandal_1707> fix keyspill
[21:25:41 CEST] <BtbN> I'm thinking about putting that into an entirely seperate filter
[21:25:56 CEST] <BtbN> As it's also usefull without chroma/colorkey
[21:26:15 CEST] <durandal_1707> ok
[21:26:24 CEST] <JEEB> durandal_1707: the idea wasn't really to be faster than swscale, rather being a more correct but still fast thing
[21:26:27 CEST] <JEEB> http://up-cat.net/p/9b8d5c6b
[21:26:39 CEST] <JEEB> it seems like there's more SIMD now
[21:27:07 CEST] <durandal_1707> but results on doom9 are paradise
[21:28:38 CEST] <durandal_1707> I test bilinear only
[21:37:11 CEST] <cone-619> ffmpeg 03Carl Eugen Hoyos 07master:98ed0716fb79: lavf/rawenc: Force one stream for hevc and m4v.
[21:40:46 CEST] <jamrial> BtbN: if the opencl maintainers didn't comment and noone else knows much about that stuff, then go ahead i guess
[21:41:11 CEST] <BtbN> I think i'll ping it once again, and directly CC the maintainer(s)
[21:43:13 CEST] <BtbN> Great, there's a name and a GnuPG Fingerprint, but no E-Mail.
[21:44:35 CEST] <nevcairiel> wasnt there a mail in the license on top of the generic opencl header
[21:44:59 CEST] <BtbN> ah, right
[21:45:03 CEST] <BtbN> 4 E-Mails
[21:46:51 CEST] Action: ubitux smiles to nevcairiel
[21:47:57 CEST] <nevcairiel> i'm sure he'll find some obscure reason to still protest
[21:48:52 CEST] <ubitux> no i don't think so :)
[22:17:42 CEST] <jamrial> nevcairiel: derp, i thought i was replying to the license header email and not a new one about typos
[22:17:53 CEST] <jamrial> yeah, typo check on fate is silly
[22:18:38 CEST] <nevcairiel> I honestly don't even care a single bit about typos in comments
[22:18:48 CEST] <nevcairiel> and having fate break because of that? please.
[23:55:23 CEST] <cone-619> ffmpeg 03Stefano Pigozzi 07master:78cc19f15ead: videotoolbox: require hardware acceleration
[00:00:00 CEST] --- Wed Sep 30 2015


More information about the Ffmpeg-devel-irc mailing list