[Ffmpeg-devel-irc] ffmpeg-devel.log.20120222

[burek]

[00:09] <overflow_0f8b> hi
[00:10] <overflow_0f8b> amiga is reloaded ! http://a-eon.com/x1000.html http://en.wikipedia.org/wiki/AmigaOne_X1000
[00:10] <overflow_0f8b> 2010 summer with programmable logic coprocessor
[00:25] <buzz_> overflow_0f8b, overpriced, underpowered.
[00:26] <overflow_0f8b> yeah overpriced, but programmable logic coprocessor ownz all instruction sets
[00:26] <buzz_> not really
[00:27] <buzz_> dave haynie doesnt think its much use.. some extra speed on the main cpu would be more useful
[00:27] <overflow_0f8b> buzz_<< your CPU is just a large macro of logic gates.
[00:28] <buzz_> are you talking about the xcore ?
[00:29] <buzz_> because that isnt a programmable logic coprocessor 
[00:29] <buzz_> also, what has any of this got to do with ffmpeg ? :)
[00:30] <buzz_> well. it is a coprocessor with an instruction set - but i mean it isnt like some reprogrammable device like an fpga
[00:32] <buzz_> (afaik). anyway. david haynie (who knows more about hardware than either of us im sure). didn't think it would be much use
[00:34] <buzz_> also see: http://www.osnews.com/thread?431085
[00:34] <overflow_0f8b> well i was just thinking about the possibility of having a programmable logic coprocessor, don't know what the Xena actually does yet, i just read the page that advertised it as programmable logic coprocesor
[00:35] <buzz_> read the link i just gave then.
[00:35] <overflow_0f8b> it'd be cool to program some FPGA like thing for any signal processing function, then do it at 500MHz rate for example
[00:35] <ohsix> how do you schedule an fpga
[00:35] <ohsix> the problem is not so much its presence but it's use
[00:35] <overflow_0f8b> the program with FPGA's is, they need to boot :(
[00:36] <overflow_0f8b> unless they make them on the fly modifyable...
[00:36] <ohsix> they are on the fly modifiable, but the dev tools don't really support it, you need to do it by hand
[00:36] <overflow_0f8b> wow srsly?
[00:37] <ohsix> if you have 2 blocks, you need to test either of them, then both; if you have 16 blocks ...
[00:39] <overflow_0f8b> can you point out some low level xilinx hacking resources?
[00:40] <overflow_0f8b> i'd like to build an advanced system that is self modifying
[00:41] <buzz_> fpga arcade .. minimig
[00:42] <buzz_> minimig is open so you can get some info from it
[00:43] <buzz_> http://code.google.com/p/minimig/
[01:01] <overflow_0f8b> burek<< wow thanks that looks a great start
[01:01] <overflow_0f8b> i'm gonna build a mini amiga
[03:01] <CIA-17> ffmpeg: 03Ronald S. Bultje 07master * r3ab9a2a557 10ffmpeg/libavcodec/ (7 files in 3 dirs): 
[03:01] <CIA-17> ffmpeg: rv34: change most "int stride" into "ptrdiff_t stride".
[03:01] <CIA-17> ffmpeg: This prevents having to sign-extend on 64-bit systems with 32-bit ints,
[03:01] <CIA-17> ffmpeg: such as x86-64. Also fixes crashes on systems where we don't do it and
[03:01] <CIA-17> ffmpeg: arguments are not in registers, such as Win64 for all weight functions.
[03:01] <CIA-17> ffmpeg: 03Martin Storsjö 07master * ra5c50913a8 10ffmpeg/libavformat/mov.c: (log message trimmed)
[03:01] <CIA-17> ffmpeg: movdec: Adjust keyframe flagging in fragmented files
[03:01] <CIA-17> ffmpeg: For video, mark the first sample in a trun which doesn't have the
[03:01] <CIA-17> ffmpeg: sample-is-non-sync-sample flag set as a keyframe.
[03:01] <CIA-17> ffmpeg: In particular, the "sample does not depend on other samples" flag
[03:02] <CIA-17> ffmpeg: isn't enough to make it a keyframe, since later frames still can
[03:02] <CIA-17> ffmpeg: reference frames prior to that one (the flag only says that that
[03:02] <CIA-17> ffmpeg: 03Paul B Mahol 07master * r08630d25a2 10ffmpeg/libavcodec/ (pngdec.c pngenc.c): 
[03:02] <CIA-17> ffmpeg: png: make .long_name more descriptive
[03:02] <CIA-17> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[03:02] <CIA-17> ffmpeg: Signed-off-by: Diego Biurrun <diego at biurrun.de>
[03:02] <CIA-17> ffmpeg: 03Aneesh Dogra 07master * r33510e86b1 10ffmpeg/libavformat/gif.c: 
[03:02] <CIA-17> ffmpeg: gif: K&R formatting cosmetics
[03:02] <CIA-17> ffmpeg: Signed-off-by: Diego Biurrun <diego at biurrun.de>
[03:02] <CIA-17> ffmpeg: 03Paul B Mahol 07master * rca085e667b 10ffmpeg/libavcodec/mlpdec.c: 
[03:02] <CIA-17> ffmpeg: mlpdec: use av_log_ask_for_sample()
[03:02] <CIA-17> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[03:02] <CIA-17> ffmpeg: Signed-off-by: Janne Grunau <janne-libav at jannau.net>
[03:02] <CIA-17> ffmpeg: 03Janne Grunau 07master * re48a70e6da 10ffmpeg/avplay.c: 
[03:02] <CIA-17> ffmpeg: avplay: fix -threads option
[03:02] <CIA-17> ffmpeg: The AVOptions based default to threads auto in 2473a45c8
[03:02] <CIA-17> ffmpeg: works only if avplay does not use custom option handling
[03:03] <CIA-17> ffmpeg: for -threads.
[03:03] <CIA-17> ffmpeg: CC: <libav-stable at libav.org>
[03:03] <CIA-17> ffmpeg: 03Alex Converse 07master * rb0f29db5c2 10ffmpeg/ (18 files in 5 dirs): Mark mutable static data const where appropriate.
[03:03] <CIA-17> ffmpeg: 03Diego Biurrun 07master * r24d150b773 10ffmpeg/LICENSE: Clarify licensing information about files borrowed from libjpeg.
[03:03] <CIA-17> (38 lines omitted)
[03:58] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r8a08503b78 10ffmpeg/libavcodec/pngdec.c: 
[03:58] <CIA-17> ffmpeg: pngdec: Print error if signature is wrong.
[03:58] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[03:58] <CIA-17> ffmpeg: 03Paul B Mahol 07master * rd00bc6a8fd 10ffmpeg/libavcodec/ (r210enc.c y41penc.c): 
[03:58] <CIA-17> ffmpeg: r210enc and y41penc: don't set same pkt->size twice
[03:58] <CIA-17> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[03:58] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:38] <durandal_1707> owner of x86_64-linux-gnu-gcc-4.6 N-38118-gd00bc6a should finally rsync fate samples
[04:53] Action: michaelni thinks bcoudurier is owner of x86_64-linux-gnu-gcc-4.6, but iam not 100% sure
[08:34] <Zeranoe> Could anyone explain to me the package stagefright-h264? Should it be included in the Windows builds? Is it just a different h264 decoder?
[08:43] <Zeranoe> Is it only really helpful on Android? That seems to be the main search result.
[08:45] <ohsix> it's from android
[11:02] <CIA-17> ffmpeg: 03Nicolas George 07master * r4217dfe87b 10ffmpeg/libavfilter/af_pan.c: 
[11:02] <CIA-17> ffmpeg: af_pan: remove dual double/int storage of gain.
[11:02] <CIA-17> ffmpeg: libswresample takes care of that now.
[11:47] <CIA-17> ffmpeg: 03Nicolas George 07master * r9cbf17e9af 10ffmpeg/ (4 files in 2 dirs): lavc: introduce av_get_pcm_codec.
[11:47] <CIA-17> ffmpeg: 03Nicolas George 07master * r534a82a3fd 10ffmpeg/libavdevice/lavfi.c: lavd/lavfi: support all sample formats.
[11:57] <ubitux> michaelni: any reason tests/copycooker.sh is not used?
[11:57] <ubitux> it seems we don't have any copy tests
[11:58] <ubitux> mmh my bad, there are actually a few
[13:11] <kriegerod> is it possible to pull from audio filtergraph the needed number of samples, but not that many as there was in pushed AVFrame?
[13:11] <kriegerod> oh, sorry, wrong chat, that's for #ffmpeg
[13:11] <ubitux> i was talking about something like that recently with saste
[13:12] <ubitux> kriegerod: no it's fine here :)
[13:12] <kriegerod> so what was the result?
[13:12] <ubitux> actually, it would be nice to have an audio filter to act as buffering
[13:12] <ubitux> something like adding fixedbuffer=512 at the end of the filtergraph
[13:13] <kriegerod> ubitux, i could implement it if you direct me
[13:13] <kriegerod> seems to me like just creating AVFifo inside the filter instance
[13:14] <ubitux> yes, and you might also need to pad with silence at the end
[13:14] <kriegerod> i think better name is 'setframesize'
[13:14] <kriegerod> well, "asetframesize"
[13:15] <ubitux> it was just a random guess
[13:15] <ubitux> this could be discussed on ffmpeg-devel, feel free to do it
[13:16] <ubitux> this would have various usages btw
[13:16] <ubitux> i was thinking of auto inserting such filter if some filters require a fixed size
[13:16] <ubitux> (for example fft filters)
[13:16] <ubitux> also, it could simplifies the usage of the API
[13:17] <kriegerod> yep
[13:17] <kriegerod> i thought about such filter to avoid avfifo management inside app
[13:17] <kriegerod> could you advise to start from another audiofilter as a template?
[13:17] <ubitux> i encourage you to discuss this on ffmpeg-devel ml, i'm sure Nicolas and saste would like to comment
[13:17] <ubitux> mmmh
[13:18] <ubitux> lemme see
[13:19] <ubitux> af amerge changes the number of samples, so you might want to have a look
[13:19] <ubitux> but it's a complicated filter
[13:19] <ubitux> also, af volume is a nice simple filter which deal with various sample format
[13:26] <ubitux> you will need to implement request_frame() btw (to tell if the current samples window is complete, and thus if there is data to pull)
[13:27] <ubitux> i mean poll_frame sorry
[13:28] <ubitux> i don't think any audio filter use it atm btw
[13:28] <ubitux> actually, abuffer does
[14:53] <kriegerod> is that right that for audio, AVFilterBufferRef.format holds value from enum AVSampleFormat?
[14:54] <ubitux> iirc, yes
[15:21] <ubitux> heh, that filter could also fix a bug i have in my WIP -af
[15:22] <ubitux> we could use it to request the output stream frame size if codec has not a variable frame size cap
[15:40] <kriegerod> in avfilter.h, in comment on poll_frame, there's noted "Output video pads only". Is that correct?
[15:40] <kriegerod> also on request_frame
[15:40] <ubitux> since they're in use in audio filters, i guess it's not accurate anymore
[15:41] <ubitux> a leftover from where there was only video support in lavfi i guess
[15:43] <CIA-17> ffmpeg: 03Nicolas George 07master * rf16c950f32 10ffmpeg/libavutil/common.h: lavu: document av_reverse.
[15:55] <ubitux> can't we do something about the assert warnings on openbsd?
[15:56] <ubitux> and maybe the external declarations with icc
[15:58] <durandal_1707> ubitux: what exact warnings?
[15:59] <ubitux> http://fate.ffmpeg.org/log.cgi?time=20120222094717&log=compile&slot=x86_32-linux-gnu-icc-2011.4.191
[15:59] <ubitux> http://fate.ffmpeg.org/log.cgi?time=20120222045648&log=compile&slot=x86_64-openbsd4.8-gcc4.2
[15:59] <ubitux> oh and the fpic for mingw:
[15:59] <ubitux> http://fate.ffmpeg.org/log.cgi?time=20120221211007&log=compile&slot=x86_64-debian-mingw32-gcc-4.6
[15:59] <ubitux> these configs are flooded with unwanted warnings
[16:01] <michaelni> ubitux, dont remember why copycooker isnt used, maybe something didnt work or it was just unfinished+forgotten 
[16:01] <ubitux> ok
[17:14] <overflow_0f8b> is this the simplest and best way getting the fractional part of a float ? ;/           (a).f = ( abf ) - floorf( abf );
[18:56] <Plorkyeran> modf
[20:34] <overflow_0f8b> Plorkyeran<< but floatingpoint division is slow
[20:35] <overflow_0f8b> #define fractf( f )    ( copysign( fmodf( fabsf( f ), 1.0 ), f ) )
[20:35] <overflow_0f8b> #define fractf( f )    ( fmodf( f, copysign( 1.0, f ) ) )
[20:35] <overflow_0f8b> also...
[20:36] <Plorkyeran> what does modf have to do with floating point division
[20:36] <Plorkyeran> modf != fmodf
[20:39] <overflow_0f8b> what does modf have to do with my floating point rounding?
[20:39] <overflow_0f8b> i don't see it in tgmath.h btw
[21:11] <overflow_0f8b> Plorkyeran<< lol thanks, i was doing the exact same thing as modf ;/
[21:12] <overflow_0f8b> it breaks down a float to an integer and fractional part
[21:23] <overflow_0f8b> hopefully it won't start floatingpoint dividing ...
[21:27] <overflow_0f8b> fi = modff( abf, &abf);
[23:00] <durandal_1707> ffplay seems to ignore pkt->duration
[23:15] <michaelni> durandal_1707, probably best if you mail marton or report it on trac
[23:19] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r43b1943a55 10ffmpeg/ (doc/eval.texi libavutil/eval.c tests/ref/fate/eval): 
[23:19] <CIA-17> ffmpeg: eval: Add taylor series evaluation support.
[23:19] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:19] <CIA-17> ffmpeg: 03Michael Niedermayer 07master * r094673ff1b 10ffmpeg/libavutil/eval.c: 
[23:19] <CIA-17> ffmpeg: eval: support 3 parameter functions.
[23:19] <CIA-17> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:00] --- Thu Feb 23 2012