[Ffmpeg-devel-irc] ffmpeg-devel.log.20121219

burek burek021 at gmail.com
Thu Dec 20 02:05:02 CET 2012


[00:00] <Tjoppen> ubitux: where is it from?
[00:00] <ubitux> deringing in pp
[00:00] <ubitux> libpostproc/postprocess_template.c
[00:01] <ubitux> around L1350
[00:02] <Tjoppen> hm
[00:02] <Tjoppen> it's some kind of kernel for deringing?
[00:03] <ubitux> looks like some kind of weigthed histogram or something to me
[00:03] <Tjoppen> this feels worthy of posting somewhere
[00:03] <ubitux> but i really have no idea how that work
[00:04] <Tjoppen> it's ANDing the s[y] together
[00:04] <Tjoppen> ah, and shifts them back. hrmm
[00:04] <ubitux> http://24.media.tumblr.com/tumblr_mamw0rbkSE1qbsw6yo1_400.gif
[00:04] <ubitux> magic
[00:05] <ubitux> saste: so no problem with removing the auto clipping from gradfun options?
[00:05] <Tjoppen> ok, so the (t<<1) & (t>>1) bit basically checks for adjacent pixels which are brighter than average
[00:05] <Tjoppen> for each pixel
[00:08] <Tjoppen> let's see who the hell wrote that mess :)
[00:10] <Tjoppen> 2e212618 postproc/postprocess_template.c               (Michael Niedermayer  2001-10-26 00:35:41 +0000 1448)            t |= (~t)<<16;
[00:10] <Tjoppen> not big surprise
[00:30] <ubitux> we lost michael :(
[00:33] <kierank> cia got him
[00:33] <ubitux> :(
[00:37] <saste> ubitux: ok with removing the clipping to me
[00:37] <ubitux> ok, i'll push in a moment then
[00:37] <saste> s/to me/for me/
[00:49] <cbsrobot> Tjoppen: ping
[00:49] <cbsrobot> your mxf patch breaks fate
[00:51] <Tjoppen> oh noes
[00:53] <cone-798> ffmpeg.git 03Michael Niedermayer 07master:c4a32d92fee3: ffmpeg: Fix shortest with libx264
[00:55] <Tjoppen> it's late so I'll look at it tomorrow
[01:02] <ubitux> michaelni: btw, since you are working looking on libx264, there is a valgrind problem i wasn't able to figure out
[01:02] <ubitux> are you interested?
[01:03] <ubitux> ’ x4->out_pic.quality is sometimes undefined so it leads to uninitialized read in ffmpeg.c (for stats)
[01:03] <ubitux> i don't really understand why it's not init
[01:03] <michaelni> how to reproduce ?
[01:04] <ubitux> valgrind ./ffmpeg_g -f lavfi -i testsrc -frames:v 2 -c:v libx264 -f null -
[01:05] <ubitux> i tried something like http://pastie.org/5549927
[01:05] <ubitux> but i'm not sure to understand why it's needed, and i got another problem iirc
[01:57] <michaelni> ubitux, my valgrind here does not work with x264, later uses some instruction that valgrind cant handle
[02:00] <ubitux> mmh, no libx264 option to disable optim or sth?
[02:00] <ubitux> maybe a simpler profile could help
[02:02] <ubitux> this is what i have: http://pastie.org/private/kvdstjysewqb1bmnngq
[02:26] <michaelni> ubitux, here valgrind gives for this: vex amd64->IR: unhandled instruction bytes: 0x66 0xF 0x38 0x41 0xC0 0x66 0xF 0x7E
[02:27] <michaelni> valgrind-3.7.0
[02:31] <drv> 660F3841C0 is phminposuw xmm0,xmm0
[02:36] <ubitux> p`pwgen` xmm0,xmm0
[02:37] <cone-798> ffmpeg.git 03Paul B Mahol 07master:dea9da5f1650: qcelpdata: correct start bit for cbgain[0] in qcelp_rate_octave_bitmap[]
[02:37] <ubitux> p`pwgen -N1` actually
[02:47] <drv> hehe, the newer SSE instructions do look a lot like line noise :)
[02:58] <iive> ph min pos uw, finds the minimus position in unsigned words? (just guessing from the op name)
[03:10] <ubitux> michaelni: did you try -cpuflags -sse4 or something?
[03:11] <ubitux> (not sure if that get transmitted to libx264 though)
[03:17] <durandal_1707> does anyone have als spec?
[03:17] <Compn> check mikes mirror page ?
[03:18] <durandal_1707> it is in doc and not full
[03:18] <durandal_1707> i want pdf one which is sold
[03:18] <cone-798> ffmpeg.git 03Clément BSsch 07master:dd8a76ab080d: lavfi/gradfun: do not increment DC pointer for odd values.
[03:18] <cone-798> ffmpeg.git 03Clément BSsch 07master:49de902a1e07: lavfi/gradfun: fix rounding in MMX code.
[03:18] <cone-798> ffmpeg.git 03Clément BSsch 07master:60ba9a9a88fe: lavfi/gradfun: fix dithering in MMX code.
[03:18] <cone-798> ffmpeg.git 03Clément BSsch 07master:63e1fc25884f: lavfi/gradfun: remove rounding to match C and SSE code.
[03:18] <cone-798> ffmpeg.git 03Clément BSsch 07master:3bdd70fc661f: lavfi/gradfun: reduce up limit for threshold.
[03:18] <cone-798> ffmpeg.git 03Clément BSsch 07master:ca83e6ce266b: fate: re-enable lavfi gradfun filter test.
[03:18] <cone-798> ffmpeg.git 03Clément BSsch 07master:526e2415bd1e: lavfi/gradfun: support named options.
[03:18] <cone-798> ffmpeg.git 03Clément BSsch 07master:fea6f5e14d1c: doc/filters: reformat and rework gradfun.
[03:19] <Compn> durandal_1707 : hmm, maybe i can locate it
[03:20] <Compn> which one is als ?
[03:21] <Compn> MPEG-4 Audio Lossless Coding (ALS)
[03:21] <ubitux> i think it's defined in the 1.4k pages from the audio mpeg4 part
[03:21] <Paranoialmaniac> yes
[03:23] <cone-798> ffmpeg.git 03Clément BSsch 07master:de2022c5dd00: lavfi/gradfun: small align cosmetics.
[03:25] <michaelni> durandal_1707, ask thilo, he should have a als spec
[03:32] <durandal_1707> looks like native opus decoder is stalled
[04:21] <cone-798> ffmpeg.git 03Michael Niedermayer 07master:1166fc07a20e: utils: fix duration calculation for strange_duration_example.ts
[04:26] <ubitux> michaelni: why is copyAhead=9 for dering, while it seems to be reading up to 10 values?
[04:37] <michaelni> ubitux, its a long time ago since i wrote this code, but dering reads a 1 pixel border on all sides of a 8x8 block
[04:38] <michaelni> the copy ahead probably just counts the 1 ahead of the bllock
[04:38] <ubitux> mmh, ok
[04:38] <ubitux> thx
[07:15] <wm4> ubitux: hilariously outdated, linked from the ffmpeg documentation page: http://wiki.multimedia.cx/index.php?title=FFmpeg_filter_HOWTO
[07:16] <wm4> though I guess there are worse offenders
[07:16] <wm4> like this (date 2004, last updated 2009): http://www.inb.uni-luebeck.de/~boehme/using_libavcodec.html
[09:35] <cone-267> ffmpeg.git 03Stefano Sabatini 07master:ca3d786227ad: doc/texi2pod: rework man inclusion logic
[09:35] <cone-267> ffmpeg.git 03Stefano Sabatini 07master:461716d3bfe3: doc: use a common authors.texi file blurb
[09:35] <cone-267> ffmpeg.git 03Stefano Sabatini 07master:c2ac6477106a: doc: rework synopsis section for ff* tools
[09:35] <cone-267> ffmpeg.git 03Stefano Sabatini 07master:9d6517aa674b: doc: extend "See Also" chapter rendering for ff* tools
[09:35] <cone-267> ffmpeg.git 03Stefano Sabatini 07master:9b323a9b9528: doc: extend "See Also" chapters in component manuals
[09:35] <cone-267> ffmpeg.git 03Stefano Sabatini 07master:c5d7c7d33bb9: doc: extend "See Also" chapters in library manuals
[10:01] <cone-267> ffmpeg.git 03Carl Eugen Hoyos 07master:ec4a084581f7: Show underscores in av_get_codec_tag_string().
[11:51] <cone-267> ffmpeg.git 03Stefano Sabatini 07master:022e118e3fe1: lavfi/setpts: enable debug logs, and improve/extend debug messages
[12:35] <Matias^> hi
[12:36] <Matias^> anybody here ever tried to use -use_wallclock_as_timestamps ?
[12:38] <Matias^> i've got a feeling it doesn't work with the isml muxer/segmenter
[12:58] <ubitux> wm4: yeah, except the official documentation, everything is pretty crambling
[12:58] <wm4> ubitux: better delete this shit
[13:00] <cone-267> ffmpeg.git 03Clément BSsch 07master:80a9297b711b: doc: remove mention of fifo filter in the introduction.
[13:00] <ubitux> yepee Nicolas is alive.
[14:38] <cone-267> ffmpeg.git 03Diego Biurrun 07master:523c7bd23c78: misc typo, style and wording fixes
[14:38] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:af804dbe9ed0: Merge commit '523c7bd23c781aa0f3a85044896f5e18e8b52534'
[14:44] <cone-267> ffmpeg.git 03Janne Grunau 07master:73ad2c2fa725: h264: increase dist_scale_factor for up to 32 references
[14:44] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:ed2d7d5868a4: ff_h264_direct_ref_list_init: fix B slice check.
[14:44] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:e84194f06024: Merge commit 'ed2d7d5868a4a5d914f1e5488d63ea696a3b2937'
[15:49] <cone-267> ffmpeg.git 03Janne Grunau 07master:f1d8763a02b5: mpegvideo: allocate scratch buffers after linesize is known
[15:49] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:642a655f7d9a: Merge commit 'f1d8763a02b5fce9a7d9789e049d74a45b15e1e8'
[15:58] <ubitux> git-rerere - Reuse recorded resolution of conflicted merges
[15:58] <ubitux> michaelni: do you use this? :))
[16:02] <mateo`> sounds like a powerful dark magic command :]
[17:02] <cone-267> ffmpeg.git 03Paul B Mahol 07master:d85854b2691f: pvfdec: drop pointless casting of buffer in ff_get_line()
[17:59] <cone-267> ffmpeg.git 03Clément BSsch 07master:a1d14336d31c: pp: fix a few typo in the template.
[21:20] <cone-267> ffmpeg.git 03Paul B Mahol 07master:76877beadac9: au: return meaningful error codes
[21:51] <durandal_1707> michaelni: what is status of ffv1/nut standardization ?
[21:52] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:25a0af51da5f: h264: factor get_pixel_format() out
[21:52] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:971c469a0d01: h264: split flush_change() out of flush_dpb()
[21:52] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:afc03268b452: h264: move list_count and current_slice reset to flush_change()
[21:52] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:8ea4a5533f4d: h264: move h264_set_parameter_from_sps() call up
[21:52] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:04a530f7d3fc: ffmpeg_filter: use frame dimensions instead of decoder context dimensions.
[21:52] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:7e6c68394f23: ffmpeg_filter: use frame instead of decoder context pixel format
[21:52] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:c2dfb1e37cc7: h264: support frame size changes with multi threading
[21:52] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:e33811bd2686: h264: remove reschange blocking
[21:52] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:af0fafcb6f08: h264: decode_update_thread_context() copy parameter sets before using them
[21:52] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:ca4dd3810e20: h264: use must_reinit to simplify code
[21:54] <michaelni> durandal_1707, nut should be fine, ffv1 (pre v3) should be fine too v3 is something i want to work on but always have other things to do 
[21:57] <durandal_1707> michaelni: i mean about sending it to org like what happened to opus
[21:59] <michaelni> we should
[22:00] <michaelni> but iam busy
[22:01] <michaelni> if you and peter/dericed want to submit it to some org i do support this.
[22:03] <cone-267> ffmpeg.git 03Janne Grunau 07master:9e696d2e5ffa: h264: support frame parameter changes during frame-mt
[22:03] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:f61498e1dfba: Merge commit '9e696d2e5ffaa763c564682ec18c3b51b3e5fccc'
[22:24] <cone-267> ffmpeg.git 03Anton Khirnov 07master:10bca66101b7: cmdutils: add a macro to simplify grow_array() calls.
[22:24] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:330983b728cc: Merge commit '10bca66101b79954512cd9d8ee20c3bc4513adf3'
[22:30] <cone-267> ffmpeg.git 03Anton Khirnov 07master:a01c02bcff75: cmdutils: split parse_option().
[22:30] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:465c7de7924d: Merge commit 'a01c02bcff755a7891f777706b27452203aac71e'
[22:46] <cone-267> ffmpeg.git 03Anton Khirnov 07master:8c9af5b2051b: cmdutils: add a commandline pre-parser.
[22:46] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:33f9b2436aab: Merge commit '8c9af5b2051b9927f845c7afdfeb30b82670ee77'
[22:52] <cone-267> ffmpeg.git 03Anton Khirnov 07master:0ec1642b60a5: cmdutils: don't log an error in opt_default().
[22:52] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:61bc3d80e9d8: Merge commit '0ec1642b60a5b1a4d913e2b9e3899173dff77e7a'
[23:07] <ubitux> does anyone know if jon plans to add more recent versions to his ppa?
[23:07] <ubitux> 'case 0.10 is starting to get old
[23:07] <ubitux> cause*
[23:09] <michaelni> ubitux, no idea, did you ask jon ?
[23:10] <ubitux> nope, i was just wondering
[23:10] <ubitux> is he on irc?
[23:17] <cone-267> ffmpeg.git 03Clément BSsch 07master:cb0881c7948a: ffmpeg: use GROW_ARRAY in opt_map_channel.
[23:50] <J_Darnley> Does libavfilter know colour space/pixel formats when initialising filters yet?
[00:00] --- Thu Dec 20 2012


More information about the Ffmpeg-devel-irc mailing list