[FFmpeg-devel-irc] IRC log for 2010-06-07

irc at mansr.com irc at mansr.com
Tue Jun 8 02:00:38 CEST 2010


[00:10:51] <BBB> Yuvi: I think I fixed about most of the MC problem, thanks for helping yesterday, I'll see if I can hook up ff_mc_emulated_edge() or other parts of h264.c
[00:11:02] <BBB> it actually looks a little good now
[00:22:47] <BBB> Dark_Shikari: so I suppose what you meant is that I should implement something like rv34_mc in rv34.c?
[00:23:12] <BBB> (i.e. a tiny wrapper around emulated_edge_mc + dsp funcs for moving)
[00:24:02] <Dark_Shikari> BBB: I would think so, rv40 mc is very similar to vp8
[00:24:27] <BBB> it looks almost understandable...
[01:08:01] <BBB> Dark_Shikari: so h264 mc is done with the last 2 bits being "behind the comma", right?
[01:08:17] <Dark_Shikari> yes
[01:08:59] <BBB> Dark_Shikari: do we have code (or do we need special code) if vp8 uses the last 3 bits for "behind the comma" stuff?
[01:13:34] <Dark_Shikari> BBB: er.... vp8 uses 2 bits
[01:13:49] <BBB> no
[01:13:57] <Dark_Shikari> what do you mean no, it uses qpel
[01:14:10] <BBB> chapter 18 of the spec says it uses 3 bits
[01:14:21] <Dark_Shikari> That's because it's trying to confuse you.
[01:14:24] <BBB> oh
[01:14:34] <Dark_Shikari> They're trying to explain it in such a way that chroma gets integer positions.
[01:14:50] <Dark_Shikari> chroma uses an 8thpel filter, both in h264 and in vp8
[01:15:58] <BBB> you're probably right...
[01:17:16] <BBB> 8pel is chroma_op[] all over?
[01:17:35] <BBB> ok, thanks
[02:31:55] <Yuvie> BBB: excellent
[02:32:05] <Yuvie> Dark_Shikari: guess not, -cqp 0 is waht we decided on?
[02:34:23] <Dark_Shikari> I guess?
[02:34:26] <Dark_Shikari> it matches x264
[02:34:40] <Dark_Shikari> though it's not consistent with dirac's qscale
[02:35:03] <Yuvie> does the libx264 wrapper use qscale at all or is all cqp?
[02:35:22] <Dark_Shikari> not quite sure
[02:35:27] <Dark_Shikari> chekc it
[02:36:22] <Yuvie> looks like it's all cqp
[07:12:06] <KotH> salve
[07:18:15] <av500> moin
[07:24:06] <mru> moroning
[07:24:25] <thresh> moroning, indeed
[07:24:43] <thresh> @ andorra ;-)
[07:52:31] <lu_zero> morening
[07:52:55] <siretart> god morgon
[07:59:00] <kshishkov> hej (:
[08:12:31] <janneg> mru: any special encoder settings? bitrate?
[08:12:47] <Dark_Shikari> what are you encoding them as, h264?
[08:12:52] <mru> mpeg4
[08:13:24] <Dark_Shikari> tested that its fast enough?
[08:13:31] * Dark_Shikari makes some comment about FLV and my patch ;)
[08:13:43] <mru> should be
[08:14:24] <mru> it's 900x720
[08:15:55] <Dark_Shikari> I wonder if you could do it with 4 instead of 6
[08:16:40] <mru> 4 what?
[08:16:43] <mru> screens?
[08:16:46] <janneg> we have six displays so it's moot
[08:18:02] <Dark_Shikari> ah k
[08:18:11] <Dark_Shikari> those are low res screens
[08:18:22] <mru> 1280x1024
[08:18:38] <mru> but we can't decode at that resolution
[08:19:29] <av500> and 3 of the 6 lcds dont like the BB 1280x1024 timing
[08:20:23] <mru> av500: could you hack the edid of the displays to indicate their position?
[08:20:26] <Dark_Shikari> mru: you could, with my flv patch.  but of course you'd have to rerender, blah blah
[08:20:46] <mru> our beagles are only 600MHz
[08:20:53] <av500> mru: hack? as in reprogramm the i2c eeprom?
[08:20:58] <mru> yeah
[08:21:03] <Dark_Shikari> mru: not counting render time, the ipad does 1024x768 at over 60fps
[08:21:11] <av500> mru: no idea, never looked into it
[08:21:15] <Dark_Shikari> it's the render/etc that makes it shit itself
[08:21:27] <Dark_Shikari> either way, 900x720 is pretty nice.
[08:21:30] <av500> also, their are now moving along a german highway, not easy to reprogramm them from here...
[08:21:52] <mru> Dark_Shikari: we probably spend less time in display
[08:21:54] <av500> Dark_Shikari: cant u render frame N, while N+1 decodes?
[08:22:07] <Dark_Shikari> av500: it's not pipelined
[08:22:16] <Dark_Shikari> we don't control how their system works
[08:22:32] <av500> but u control the decode, no? cant u thread it?
[08:22:32] <mru> and I assume you want lowest possible latency
[08:22:41] <Dark_Shikari> av500: single core
[08:22:56] <av500> and "render" is in sw?
[08:23:44] <av500> single core still allows 2 threads, no? if the render sleeps/waits on a hw bit, the other thread could decode, no?
[08:24:03] <mru> omapfbplay!
[08:24:09] <Dark_Shikari> av500: it's in hw
[08:24:15] <Dark_Shikari> but it's a system call
[08:24:21] <Dark_Shikari> and no, it doesn't wait
[08:24:27] <Dark_Shikari> another thread, in the system , is _actively_ doing something
[08:24:47] <av500> so cpu time while "rendering" is just lost?
[08:24:50] <Dark_Shikari> apparently
[08:24:55] <av500> great!
[08:25:04] <mru> how much time roughly is that?
[08:25:42] <Dark_Shikari> 6-8ms per frame
[08:26:04] <mru> and decode?
[08:26:21] <Dark_Shikari> 15-20ms high motion, 6-9ms low motion
[08:26:32] <Dark_Shikari> spikes up to 40ms for things like keyframes
[08:26:37] <Dark_Shikari> (doesn't matter in your case, you're not low latency)
[08:28:02] <mru> not in the slightest
[08:28:09] <mru> we buffer 200MB of decoded video
[08:28:14] <Dark_Shikari> lol
[08:29:36] <janneg> grrr, I don't find the define limiting swscale to 2048 pixels
[08:30:15] <siretart> janneg: have the xiph people approached you regarding the native vorbis encoder in ffmpeg trunk?
[08:30:21] <mru> janneg: VOFW
[08:30:57] <KotH> janneg: how does the rendering goe?
[08:31:05] <KotH> janneg: does it look like you'll finish in time?
[08:37:53] <janneg> a good name. I've seen that but it wasn't the problem
[08:38:17] <janneg> as it's aleready increased to 5120 for ARCH_X86
[08:38:31] <av500> yep
[08:38:47] <av500> you cant play the 2700x1400?
[08:38:49] <mru> try increasing it anyway
[08:39:40] <janneg> KotH: just in time, I expect the last frames to be finished early tomorrow
[08:40:19] <av500> janneg: btw, nobody will notice a part of a scene missing - if needed
[08:40:31] * mru would
[08:40:58] <janneg> mru: it works now. problem was -crop before -i
[08:41:06] <mru> :-)
[08:41:26] <mru> weren't those replaced with avfilters?
[08:41:39] <mru> or are you running an OLD VERSION???
[08:41:39] <janneg> av500: or a single frame upscaled from 1080p
[08:41:49] <av500> janneg: yep
[08:42:14] <janneg> as far as I can say they use the crop avfilter
[08:42:17] <av500> janneg: or a picture of BBB instead of BBB...
[08:42:30] <janneg> I use a checkout from yesterday
[08:42:54] <mru> hmm, maybe it was only -pad* that were killed
[08:42:58] * lu_zero books now the ticket&hotel
[08:43:13] <mru> lu_zero: when do you arrive?
[08:43:22] <lu_zero> _now_ I got confirmation I'm not necessary here in Turin
[08:43:31] <av500> mru: pad only iirc
[08:43:36] <lu_zero> mru: if the booking system doesn't glitch tomorrow morning
[08:43:54] <mru> cool
[08:44:12] <mru> booth setup starts at 2pm
[08:44:22] <lu_zero> what time you'll arrive?
[08:44:43] <lu_zero> Tegel is the right airport is it?
[08:44:43] <mru> noon-ish
[08:44:48] <mru> tegel is good
[08:44:51] <mru> it's the close one
[08:44:54] <lu_zero> good
[08:45:22] * kshishkov prefers trains
[08:45:39] * kshishkov prefers trains unless in Ukraine
[08:45:47] <mru> http://www.flickr.com/photos/av500/4673981263/
[08:45:48] <av500> nice steam trains here too :)
[08:46:07] <kshishkov> we have one here in Schlossgarten
[08:46:24] <mru> a steam train?
[08:46:29] <mru> in the park?
[08:46:36] <kshishkov> and I missed one in Stockholm last year on 12th of April IIRC
[08:46:40] <kshishkov> yes
[08:47:17] <kshishkov> http://commons.wikimedia.org/wiki/Category:Schlossgartenbahn_Karlsruhe
[08:47:49] <kshishkov> it's really a disguised diesel engine IIRC
[08:48:34] <av500> kshishkov: no, one is a real steamer
[08:48:53] <kshishkov> I believe you
[08:50:02] <wbs> speaking of stockholm, /me is there today
[08:50:17] <mru> wbs: good for you
[08:50:20] <mru> stockholm is nice
[08:50:33] <kshishkov> s/nice/very nice/
[08:50:48] <mru> at least this time of year
[08:50:54] <wbs> yeah, it's ok, except that I have to meet people in the "slipsnisse" category
[08:50:58] <mru> november-march is not so nice
[08:51:09] <mru> wbs: not good...
[08:51:13] <kshishkov> why? it's fantastic in winter
[08:51:21] <mru> stockholm, no
[08:51:27] <mru> it's miserable in winter
[08:51:31] <KotH> kshishkov: winter is never fantastic in cities
[08:51:47] <mru> some cities are better than others
[08:52:06] <mru> the climate in stockholm makes it awful in winter
[08:52:30] * KotH has never seen a city where winters are fun
[08:52:38] <KotH> though, st.gallen comes pretty close
[08:52:55] <mru> it has to be dry to have chance
[08:53:02] <mru> stockholm is not dry
[08:53:02] <KotH> but only because the city is so narrow and long, that you're within a few minutes walking deep in the forest
[08:53:11] <lu_zero> mru: Chiba is wonderful in winter
[08:53:14] * kshishkov knows a few towns that looks more or less decent only in winter
[08:53:27] <kshishkov> lu_zero: tell that KotH instead
[08:53:29] <lu_zero> (ok, it felt more or less like my local early spring...)
[08:53:50] * KotH has never been in .jp in winter
[08:54:02] <Tjoppen> ah, winter.. we got quite a good one this year, with -28 C (-40 further inland)
[08:54:27] <mru> that's too cold
[08:54:29] <KotH> Tjoppen: better than +2C we've had here
[08:54:33] <mru> -20 is plenty cold enough
[08:55:16] <spaam> Tjoppen: norrland <3
[08:55:21] <Tjoppen> :)
[08:56:01] <kshishkov> spaam: yes, I'd like to move to Kiruna for summers
[08:56:04] <lu_zero> KotH: barring hokkaido the other islands have a quite warm climate
[08:56:17] <spaam> kshishkov: why kiruna? :)
[08:56:51] <kshishkov> spaam: because you don't have large towns further to the north
[08:57:13] <KotH> lu_zero: hokkaido in mid sept felt like .ch during summer... so it's hot enough :)
[08:57:19] <spaam> kshishkov: true :)
[08:57:32] <KotH> lu_zero: though i'd love to go to hokkaido in winter
[08:57:38] <KotH> lu_zero: the snow must be beautiful
[08:57:53] <lu_zero> I want to see it in spring
[08:58:13] <KotH> how about ffcon next spring in hokkaido? :)
[08:58:31] <lu_zero> KotH: if we want to organize that I could check
[08:58:41] <kshishkov> KotH: ok, you book us tickets though
[08:58:48] <lu_zero> so at linuxtag we need to gather sponsors
[08:59:11] <KotH> kshishkov: pah! now that you work in a capitalistic country, you earn enough to pay it yourself!
[08:59:17] <KotH> ;)
[08:59:20] <spaam> haha
[08:59:29] <lu_zero> kshishkov: given most of us is in EU we could check prices for group and fly from the same place (frankfurt I think)
[08:59:38] <kshishkov> KotH: I work in the country that gave people Socialism, you ignorant man
[08:59:50] <KotH> kshishkov: nope, that was france
[09:00:04] <KotH> kshishkov: though most of the people were germans, they lived around paris at that time
[09:00:47] <kshishkov> KotH: France gave people revolutionary terror and Socialism theoreticians lived or were born in Germany, including Marx and Lenin
[09:01:21] * lu_zero meanwhile hopes to get his ticket now
[09:01:57] <kshishkov> lu_zero: a large group of European tourists in Japan with permanently flashing cameras? That should be like some kind of revenge on them
[09:01:58] <spaam> lu_zero: good luck :)
[09:02:44] <lu_zero> got it
[09:02:48] <spaam> o/
[09:03:01] <lu_zero> kshishkov: =)
[09:04:52] <KotH> kshishkov: lol
[09:05:06] <lu_zero> they got the packs of US soldier/students getting drunk
[09:05:15] <lu_zero> so I'd call it even
[09:05:23] <KotH> kshishkov: though, you'll have a hard time beating japanese in fotographing...they even do it in their own country
[09:05:34] <KotH> lu_zero: even??
[09:05:45] <KotH> lu_zero: i think there is nothing worse than usian tourists in japan
[09:05:47] <kshishkov> KotH: just take pictures of Japanese photographing everything else :P
[09:05:53] <KotH> lol
[09:06:19] <av500> kshishkov: to accomplish what?
[09:06:37] <lu_zero> KotH: there is a special kind of Italian that's nearly worse than the all-I-know-texas-turist canon
[09:07:13] <kshishkov> lu_zero: actually, the worst tourist are Japanese, Germans and Russians
[09:07:18] <lu_zero> given I'm in a city with a _bit_ of turism it happened to experience that kind here
[09:07:19] <KotH> lu_zero: hmm? i somehow doubt it
[09:07:32] <kshishkov> lu_zero: can't beat Pisa though
[09:07:44] <lu_zero> kshishkov: as people or as place?
[09:07:59] <KotH> lu_zero: i know that italians are very outspoken, but i've never met a italian so arrogant and agnostic to any culturual difference around him like the usians in .jp
[09:08:00] <kshishkov> place, of course
[09:08:33] <KotH> kshishkov: what's bad about japanese tourists?
[09:08:55] <lu_zero> KotH: think about Berlusconi, he is the prototype of that kind
[09:08:56] <kshishkov> lu_zero: you don't have tours like "and here we produce those perversions of nature known as FIAT"
[09:09:33] <KotH> lu_zero: eh.. he is bad, yes... but not as bad as some of the usians i've met in .jp
[09:09:39] <lu_zero> kshishkov: we have a large museum with al the kind of cars
[09:09:45] <lu_zero> KotH: you are making me wonder...
[09:09:49] <KotH> lu_zero: you know the meaning of "walking dictionary"?
[09:09:49] <lu_zero> anyway
[09:10:02] <lu_zero> KotH: not in the bad declination
[09:10:27] <kshishkov> KotH: and do you know that Russian academy of sciences decided to forbid publications in English?
[09:10:30] <KotH> lu_zero: it's a japanese girlfriend, you take everywhere as a translator so that you dont have to learn japanese
[09:10:55] <KotH> lu_zero: i've met usians in .jp who've lived there for years, w/o speaking even a word of japanese
[09:11:09] <lu_zero> KotH: are you sure they aren't just timid?
[09:11:11] <KotH> kshishkov: now, that i can understand :)
[09:11:30] <KotH> lu_zero: nope, arogant enough that the enviroment around them has to accomodate for the fact that they only speak english
[09:11:38] <lu_zero> ough
[09:11:50] <mru> they do that everywhere
[09:11:55] <mru> and everybody hates them for it
[09:12:05] <KotH> lu_zero: heck.. the first thing i've got asked when i whent to the TLUG meeting was, whether i have a walking dictionary
[09:12:14] <wbs> oh how easy life would be if you could manage with knowing just one language
[09:12:17] <Dark_Shikari> mru: eh, in a lot of places it's the reverse
[09:12:19] <lu_zero> ugh
[09:12:22] <Dark_Shikari> they will tell you to speak english
[09:12:31] <Dark_Shikari> because they understand you better in english than their own language
[09:12:31] <lu_zero> my tlug meeting was more than friendly
[09:12:33] <Dark_Shikari> see: the netherlands
[09:12:39] <mru> Dark_Shikari: americans always refuse to learn the local language
[09:12:52] <Dark_Shikari> mru: in most countries, they're advised not to
[09:12:59] <Dark_Shikari> Then again, japan, wtf?
[09:13:01] <lu_zero> and there were few locals but the non-native were fully melded in the customs
[09:13:07] <Dark_Shikari> How the hell could you possibly live in japan without knowing japanese
[09:13:07] <mru> in most countries, the locals appreciate if you make an effort
[09:13:09] <Dark_Shikari> that's... not possible
[09:13:12] <mru> even if you then end up speaking english
[09:13:24] <kshishkov> France too?
[09:13:30] <mru> france is possible
[09:13:31] <Dark_Shikari> mru: true, though dutch have told me that they would much rather a foreigner speak english than even good dutch
[09:13:40] <mru> that may be
[09:13:43] <Dark_Shikari> because "good" is never good enough
[09:13:46] <Dark_Shikari> ever
[09:13:53] <mru> but I think the gesture would still be appreciated
[09:14:52] <Dark_Shikari> though, english is interesting in that it's relatively easy to learn it badly
[09:14:54] <av500> most french are quite surprised that I speak it quite well
[09:14:56] <Dark_Shikari> and hard to learn it well
[09:15:06] <Dark_Shikari> and furthermore, everyone is used to people speaking english badly
[09:15:09] <mru> easy to learn, hard to master
[09:15:17] <Dark_Shikari> which isn't true of most languages
[09:15:39] <mru> the hard part of english is the insanely large vocabulary
[09:15:50] <mru> and certain obscure but powerful grammatical constructs
[09:15:53] <Dark_Shikari> not as bad as symbol-based languages
[09:16:04] <mru> script is a different issue
[09:16:06] <Dark_Shikari> true
[09:16:07] <KotH> lu_zero: depends on the people you meet there :)
[09:16:20] <kshishkov> mru: could you tell that to your teacher in Russian?
[09:16:26] <KotH> lu_zero: i've been three times at a tlug meeting, each time the attendands were completely different :)
[09:16:27] <Dark_Shikari> though, with the large vocabulary, even native speakers are used to learning new words
[09:16:47] <KotH> lu_zero: interestingly, the mailinglist has much less fluctuation
[09:18:41] <Dark_Shikari> though english has those nasty pronounciation problems, like the terrible "th"
[09:18:53] <Dark_Shikari> though at least it's not estonian or something
[09:19:12] <kshishkov> Hungarian
[09:19:38] <kshishkov> or (P)Russian
[09:19:56] <KotH> Dark_Shikari: english is one of the languages with the most consonants
[09:20:09] <KotH> kshishkov: klingon ;)
[09:20:11] <Dark_Shikari> english??!
[09:20:13] <Dark_Shikari> hahahahahahaha
[09:20:29] <Dark_Shikari> I think you meant somewhere around kyrgystan ;)
[09:20:52] <mru> slavic languages have much more consonants
[09:21:02] <mru> swedish has more than english too
[09:21:03] <av500> +1
[09:21:15] <kshishkov> http://en.wikipedia.org/wiki/Wikipedia:IPA_for_Russian
[09:22:32] * KotH doesnt know swedish
[09:22:44] <KotH> but english has definitly more consonants than any other language i know
[09:22:51] <Dark_Shikari> just pronounce w as v
[09:22:58] <Dark_Shikari> and say "bork bork bork" every once in a while
[09:23:31] <kshishkov> that was "Swedish for Dummies from Texas, W-version"
[09:23:36] <KotH> bork bork bork
[09:23:48] <av500> børk?
[09:23:50] <mru> my swedish work list has a few words with 7 consecutive consonants
[09:23:57] <mru> av500: that would be norwegian
[09:23:58] <kshishkov> av500: that's Danish
[09:24:10] <wbs> mru: what word would that be? :-)
[09:24:14] <mru> branschstruktur
[09:24:26] <av500> branch structure?
[09:24:32] <wbs> hah, nice
[09:24:35] <av500> sounds totally swedisch
[09:24:42] <mru> and världsspråk
[09:24:57] <wbs> finnish is quite heavy on vowels, on the other hand
[09:25:03] <wbs> hääyöaie is a proper word ;P
[09:25:21] <Dark_Shikari> so is onomotopoeia
[09:25:33] <Dark_Shikari> or queueing
[09:25:35] <lu_zero> what's that?
[09:25:39] * lu_zero runs
[09:25:45] <Dark_Shikari> "ribbit ribbit"
[09:25:46] <av500> lu_zero: you stand in a line
[09:25:46] <lu_zero> I'll be back in a while
[09:25:47] <Dark_Shikari> "oink oink"
[09:25:50] <Dark_Shikari> those are onomotopoeia
[09:25:52] <kshishkov> ni!
[09:26:04] <Dark_Shikari> words which are sound like what they are supposed to mean
[09:26:27] <Dark_Shikari> Japanese is the international king of onomotopoeia
[09:26:29] <KotH> mru: if sch in swedish is pronounced the same way as in german, then it counts phonetically as only one consonant
[09:26:32] <lu_zero> I got my ticket and booking done so it's better if I condense a week in the next 6 hours =P
[09:26:34] <Dark_Shikari> they have onomotopoeia for things that aren't even _sounds_
[09:26:37] <Dark_Shikari> like "jump"
[09:26:44] <kshishkov> KotH: it's not
[09:26:55] <kshishkov> or "glitter"
[09:27:16] <lu_zero> Dark_Shikari: "zomp" is a valid onomatopoeia for a jump here
[09:27:21] <Dark_Shikari> lol
[09:27:41] <Dark_Shikari> also of course, in japanese, every single onomotopoeia is repeated.
[09:27:43] <Dark_Shikari> "kero kero"
[09:27:54] <mru> what a waste
[09:28:00] <KotH> Dark_Shikari: not true
[09:28:01] <lu_zero> a frog
[09:28:07] <lu_zero> ?
[09:28:08] <KotH> Dark_Shikari: it's true that a lot are repeated, but not all
[09:28:12] <Dark_Shikari> which ones aren't?
[09:28:18] <Dark_Shikari> lu_zero: yes
[09:29:13] <lu_zero> "cra cra" (stragely in Italian frogs and crows collide)
[09:29:26] <KotH> Dark_Shikari: "gurutte mawaru" is a standard expression
[09:29:30] * av500 guesses the cow wins
[09:29:33] <Dark_Shikari> KotH: for what?
[09:29:40] <Dark_Shikari> I can't figure that one out
[09:29:55] <KotH> Dark_Shikari: meaning "running around in a circle" or "running around the globe"
[09:30:05] <lu_zero> uhmm
[09:30:11] <Dark_Shikari> lol, they have onomotopoeia for that?
[09:30:22] <Dark_Shikari> I never cease to be amazed by japan's love of Unsound Effectsw
[09:30:25] <Dark_Shikari> -w
[09:30:45] <KotH> Dark_Shikari: only the first word is, "mawaru" is a normal word (to turn around, to circle)
[09:30:49] <Dark_Shikari> ah
[09:30:55] <Dark_Shikari> http://tvtropes.org/pmwiki/pmwiki.php/Main/UnsoundEffect etc
[09:31:17] <kshishkov> Dark_Shikari: and now you've stolen elenril's bread, shame on you!
[09:31:19] * KotH will not look at tvtropes during work ^^'
[09:31:30] <Dark_Shikari> niko niko
[09:31:40] <lu_zero> video?
[09:31:45] <mru> KotH: tvtropes is mostly SFW
[09:32:03] <Dark_Shikari> mru: not safe for time
[09:32:07] <lu_zero> eh eh
[09:32:07] <Dark_Shikari> for getting work done =p
[09:32:14] <bcoudurier> damn earthquake at manhattan beach
[09:32:34] <KotH> mru: safe for work yes...
[09:32:35] <bcoudurier> 3.6 but when you're near, you feel it :>
[09:32:58] <KotH> bcoudurier: 3.6?
[09:33:04] <av500> richter
[09:33:20] <KotH> bcoudurier: sounds like a small car driving a few meters past you
[09:33:54] <bcoudurier> yeah, caltech says 3.6 but it really shook
[09:34:24] * KotH thinks the stronges he experienced in jp was a 5.3
[09:34:43] <KotH> during supper... and nobody took notice ;)
[09:34:55] <kshishkov> they're used to earthquakes
[09:34:59] <Dark_Shikari> I only experienced one in california
[09:35:05] <Dark_Shikari> it was when there was like a 6.something on the mexican border
[09:35:16] <Dark_Shikari> we were on the top floor of the office building, it swayed for about a minute
[09:36:08] <bcoudurier> I've experienced a few in santa monica
[09:36:17] <bcoudurier> sometimes wake me up during the night :/
[09:36:19] <KotH> well.. the building i was in swayed even when a japanese truck (at most 20t) was driving on the main road, 50m away
[09:36:42] <kshishkov> don't build them from paper then
[09:37:36] <av500> swaying from truck is needed to survive earthquake I guess
[09:38:07] <kshishkov> just build your castle on a swamp
[09:38:34] <av500> you can visit castle on swamp tomorrow
[09:39:10] <kshishkov> where?
[09:40:06] <bcoudurier> 3.6 magnitude quake rattles Santa Monica Bay
[09:40:12] <bcoudurier> la times breaking news
[09:40:49] <KotH> kshishkov: well, there are two types of buildings in .jp: anyhting larger than 2 floors is made from concrete. everything else looks like a house made from matchsticks with glued on paper (litteraly)
[09:41:18] <av500> kshishkov: berlin == swamp, pick your castle...
[09:41:27] <Dark_Shikari> av500: well you can certainly tell from the smell
[09:41:37] <kshishkov> av500: I'll get there on Wednesday morning
[09:41:45] <av500> kshishkov: wednesday then :)
[09:41:57] <av500> Dark_Shikari: it literally used to be one...
[09:43:20] <elenril> KotH: i live in a house made of polystyrene ;)
[09:43:46] <KotH> elenril: people who live underneath a bridge do not count ;)
[09:44:16] * elenril kills KotH with a rusty spork
[09:44:57] <kshishkov> av500: same story with Kharkov, but it still turns into swamp after medium rain
[10:02:19] * KotH dies with a rusty spork in his stomach
[10:04:08] * elenril level up
[10:05:22] * KotH rises from the dead, gathers a ghost army and marches them to get elenril 
[10:05:49] * kshishkov reminds Koth that elenril lives in the country that built golems
[10:06:19] * KotH does not fear prehistoric robots made of clay
[10:06:31] <bcoudurier> bye guys
[10:08:07] <kshishkov> KotH: you know, golems are more powerful than ghosts or zombies
[10:08:42] <elenril> indeed
[10:08:56] <elenril> golems have are immune to everything except bashing damage
[10:09:21] * mru bashes elenril
[10:09:29] <Dark_Shikari> I prefer onix
[10:09:38] <Dark_Shikari> I like my immunity to electric types
[10:09:39] * mru prefers unix
[10:09:43] <Dark_Shikari> (wait, wrong game)
[10:09:54] <av500> mru: GNU/unix?
[10:10:02] <kshishkov> mru: you know, that's when you need a shovel to do much damage
[10:10:08] <mru> av500: llama
[10:10:24] <Dark_Shikari> plan 9?
[10:10:27] <mru> my sister has a llama
[10:10:32] <mru> don't ask me why
[10:10:48] <av500> cuteness?
[10:10:49] <Dark_Shikari> was she bitten by a moose?
[10:10:57] <kshishkov> well, quite a convenient animal for mountanous area like .ch
[10:11:45] <lu_zero> llama?
[10:11:47] <lu_zero> real one?
[10:11:55] <lu_zero> where does she live?
[10:12:02] <Dark_Shikari> as opposed to a stuffed one?
[10:12:13] <lu_zero> Dark_Shikari: that isn't a llama
[10:13:51] * elenril shoots a master spark at Dark_Shikari 
[10:14:41] <elenril> MASUTAA SUPAAKU!
[10:15:22] * kshishkov researches facepalming to perform one act of those
[10:15:34] <av500> kshishkov: m(
[10:17:35] <mru> lu_zero: yes, a real llama
[10:17:42] <mru> in the swedish countryside
[10:19:21] <lu_zero> O_O
[10:19:27] <spaam> mru: is she a farmer?  :O
[10:20:13] <kshishkov> mru: maybe it's somehow connected to the fact I saw many Inca in Stockholm last year
[10:20:39] <lu_zero> my sister shouldn't ever meet yours
[10:20:53] <kshishkov> does she want a mere pony?
[10:20:59] <mru> lu_zero: what would happen?
[10:21:07] <lu_zero> they might exchange ideas
[10:21:28] <spaam> maybe they will became bff! ;)
[10:21:32] <lu_zero> kshishkov: she does have already a forest in her house and in her room
[10:21:50] <av500> only a llama missing
[10:22:07] <kshishkov> lu_zero: I'm glad I'm the only child in my family
[10:22:21] <lu_zero> av500: eh...
[10:22:23] <lu_zero> mostly
[10:22:53] <KotH> kshishkov: think how much fun you would have had to implement those crazy ideas you had together with your brother or sister!
[10:24:00] * kshishkov thinks
[10:24:18] <kshishkov> nah, that would be rather war for resources
[10:24:19] <lu_zero> KotH: since I like plants she decided to flood me with trees...
[10:24:46] <av500> so many that you cant see the forest any more?
[10:24:58] <lu_zero> av500: let see
[10:25:10] <KotH> lu_zero: as her for chilli next time
[10:25:24] <lu_zero> I started with just some berries, a citrus and some bamboo
[10:26:00] <lu_zero> now I have 4 kind of berries, 4-5 citruses (2 died in the winter when I was abroad), a grapewine ...
[10:26:41] <lu_zero> and she is even more fond of animals even if luckily the best she had been allowed had been cats
[10:28:00] <kshishkov> cats? bad luck
[10:29:10] <lu_zero> kshishkov: eh...
[10:33:01] <Kovensky> could've been worse; could've been dogs
[10:33:21] <Kovensky> they're evil themselves, I tell you :E
[10:34:28] <lu_zero> Kovensky: eh
[10:35:10] <lu_zero> luckily she likes better cats
[10:35:52] <kshishkov> Kovensky: and almost all evil dog breeds sound German
[10:37:30] <mru> cats are cool
[10:38:22] <elenril> http://tvtropes.org/pmwiki/pmwiki.php/Main/CatsAreMean
[10:38:38] <av500> http://hardwarebug.org/2010/02/21/cat-pictures/
[10:41:17] <janneg> mru: http://http://www.jannau.net/bbb_videowall/bbb_linuxtag2010_lt.mkv top left corner
[10:41:55] <mru> av500: first one is best
[10:42:03] <mru> those eyes...
[10:42:16] <mru> those are my sister's cats btw
[10:42:40] <janneg> there is something odd with the video, it's displayed diagonal in ffplay/mplayer
[10:42:51] <kshishkov> stride issues?
[10:42:55] <janneg> but is correct in mythtv
[10:43:04] <av500> janneg: the large one? or the cut one?
[10:43:25] <janneg> kshishkov: could be
[10:43:42] <janneg> av500: cutted one
[10:43:47] <av500> play ok here
[10:43:50] <av500> plays
[10:44:00] <mru> omapfbplay plays it ok
[10:44:06] <mru> at least with xv display
[10:44:30] <kshishkov> plays fine here even with -vo x11
[10:44:34] <merbzt1> vlc ok also
[10:44:40] <av500> janneg: double http:// is better?
[10:44:44] <mru> ffplay is fine too
[10:45:07] <lu_zero> looks fine here as well
[10:45:46] <janneg> av500: no, chromium seems to be fixed and doesn't ommit the http:// in the selection anymore
[10:45:55] <elenril> av500: that's http tunelling over http
[10:46:09] <av500> elenril: can you tunnel that over http then?
[10:46:53] <av500> janneg: do you know now the accumulated cpu time to render BBB?
[10:50:03] <janneg> av500: no if I had to guestimate: 100 days with 12 cores
[10:50:27] <av500> so 1y on an I7
[10:50:38] <av500> mru: start for next LT now :)
[10:50:51] <av500> 4k please
[10:51:21] <kshishkov> av500: we'd better do some other trick, videowalls work only once or twice
[10:51:29] <av500> ah right
[10:51:31] <av500> 3d?
[10:51:57] <janneg> we could just add displays to the videowall
[10:52:21] <mru> next time we might be able to get more powerful beagles
[10:52:34] <mru> or even something omap4 based
[10:53:09] <kshishkov> cooperate with x264 and do full-HD live streaming in H.264
[10:53:17] <kshishkov> in realtime, of course
[10:53:40] <av500> lunch!
[10:55:05] <lu_zero> kshishkov: using multiple cameras obviously
[10:55:20] <kshishkov> whatever
[10:55:52] <lu_zero> and applying different avfilters on each feed
[10:56:12] * kshishkov remembers dancing robot that they usually demonstrate every year at his university - and it's there for at least ten years still the same
[10:57:59] <lu_zero> so the show now is about its durability?
[10:58:19] <mru> and the patience of the viewers
[10:59:51] <lu_zero> every 3-5 years they got replaced so...
[11:09:49] <Compn> at least it doesnt sing 'its a small world' , i hope?
[11:10:06] <Compn> when i think of robots... i think of disney world :P
[11:10:10] <Compn> animatronics
[11:10:14] <mru> av500: hmm... beaglepack for next year?
[11:12:22] <Compn> is sita sings the blues under cc ? i wonder if there is an hd version
[11:12:58] <Dark_Shikari> yes there is
[11:13:07] <Dark_Shikari> http://media.xiph.org/video/derf/y4m/1080p/sita_sings_the_blues_1080p24.y4m.xz
[11:13:11] <Dark_Shikari> It's not very impressive though.
[11:13:13] <Dark_Shikari> Just a big long flash video.
[11:13:58] <Compn> ah good point
[11:14:06] <Compn> no one likes flash vectors :P
[11:17:51] <KotH> beside arc attack :)
[11:28:17] <CIA-92> ffmpeg: thardin * r23513 /trunk/libavformat/riff.c: Added M701 codec_tag for mpeg2video
[11:44:18] <CIA-92> ffmpeg: vitor * r23514 /trunk/libavcodec/mpegaudiodec.c:
[11:44:18] <CIA-92> ffmpeg: Optimize 1D DCT transform used in MP{1,2,3}. Makes also this function suitable
[11:44:18] <CIA-92> ffmpeg: for being moved to the shared DCT framework in the future.
[12:09:22] * kshishkov wants a plushie dead aquarium fish with a green zigzag sticking from its back
[12:12:48] <av500> mru: we'll see for next year
[12:15:38] <KotH> kshishkov: desing one, and we'll make it :)
[12:17:25] <kshishkov> KotH: I think it's easy to imagine. Especially if I also want mru to send one to Monty
[12:18:59] <KotH> lol
[12:19:59] <av500> we could at least make stickers ...
[12:20:22] <lu_zero> the xiph people will be at the linuxtag?
[12:20:43] * lu_zero today should make sure he has enough time to buy some stuff
[12:21:24] <av500> like gaffer tape?
[12:26:49] <lu_zero> av500: is that needed?
[12:27:11] <KotH> i think torches should be enough
[12:28:07] <av500> and listen to them while they roast?
[12:28:42] <janneg> I can bring gaffa tape
[12:56:49] <mru> kshishkov: ff ninja stars?
[12:57:00] <av500> http://imagebin.ca/view/m8GFEJ.html
[12:57:54] <mru> :-)
[12:58:07] <kierank> the zig zag could also choke the fish
[12:58:27] <kshishkov> or be BBQ grille for fried fish
[12:58:39] <av500> lose yourself guys
[12:59:15] <av500> yourselves?
[13:00:28] <janneg> mru: do you think we have enough space for the beast and my 37" lcd?
[13:00:47] <av500> you can put the 37" behing it
[13:00:49] <av500> d
[13:01:45] <av500> janneg: can we run a VP8 vs BINK side by side comparison on the 37"?
[13:02:06] <mru> lol
[13:02:27] <janneg> I would prefer not to move it if there's not enough space
[13:02:55] <av500> it cant fit into a suitcase?
[13:03:19] <kshishkov> probably not
[13:03:44] <KotH> get a bigger suitcase :)
[13:03:51] <av500> or a hacksaw
[13:03:58] <KotH> and superglue
[13:04:12] <av500> oled will just grow back?
[13:08:56] <lu_zero> O_O
[13:09:29] <kshishkov> av500: make OLED evolve into better display first
[13:15:35] <KotH> and soon it'll evolve far enough to conquer the world and enslave all humans
[14:26:27] <lu_zero> peloverde: your friends didn't contact me yet, did they solve their problem already
[14:26:43] <lu_zero> ?
[14:30:22] <twnqx> janneg: any idea why blender just rendomly stops every now and then?
[14:30:29] <twnqx> not terminate, just... 0% cpu
[14:30:48] <kshishkov> swapping?
[14:31:34] <twnqx>              total       used       free     shared    buffers     cached
[14:31:35] <twnqx> Mem:         16084      14256       1828          0        110       9455
[14:31:35] <twnqx> -/+ buffers/cache:       4690      11394
[14:31:35] <twnqx> Swap:            0          0          0
[14:31:35] <twnqx> no.
[14:32:43] <twnqx> strace only sees nanosleep({0, 50000000}, NULL)          = 0
[14:35:59] <kshishkov> so it just got tired and decided to sleep a bit, no worries
[14:41:32] <janneg> twnqx: no idea
[14:41:49] <av500> janneg: I thought you are done?
[14:42:32] * BBB points mru to webm issue 67
[14:43:04] * kshishkov suspect that mru does not know or cares about webm issues
[14:43:35] <BBB> it's not to fix, it's to read and have an opinion on
[14:43:42] * mru suspects kshishkov might be right
[14:43:45] <av500> New issue 67 by christian.schaller: Build system should provide pkgconfig file
[14:43:57] <mru> BBB: link?
[14:44:07] <BBB> http://code.google.com/p/webm/issues/detail?id=67
[14:45:10] <twnqx> janneg: 4 more frames
[14:45:24] <twnqx> hm, on the other hand
[14:45:38] <mru> BBB: pointless
[14:45:45] <lu_zero> give that currently all they need is -lm and optionally pthreads not sure it is _that_ necessary
[14:45:59] <mru> libvpx has no deps
[14:46:13] <lu_zero> beside -lm and pthreads
[14:46:13] * twnqx renders the last four frames at 2 in parallel
[14:46:31] <lu_zero> (as I discovered the hard way)
[14:46:45] <mru> pkgconfig doesn't handle pthreads correctly
[14:47:20] <lu_zero> well you can put it in so it outputs -pthread in --cflags and --libs
[14:47:22] <mru> -lm is a standard flag so it's safe to always add it
[14:47:39] <lu_zero> mru: if you aren't on haiku or osx somebody complained
[14:47:40] <mru> lu_zero: wrong
[14:47:57] <lu_zero> uh?
[14:48:09] <mru> the flag depends on the compiler
[14:48:32] <mru> setting it statically simply cannot work
[14:48:46] <lu_zero> you'd have to provide it somehow else
[14:49:02] <lu_zero> but in that case elf won't save you
[14:49:14] <mru> yes, and pkgconfig doesn't
[14:49:15] <lu_zero> -lmylibthatusespthreads isn't enough
[14:49:25] <lu_zero> am I wrong?
[14:50:21] <mru> in theory all code going into a threaded app can need special cflags
[14:51:17] <mru> stdio macros come to mind
[14:52:02] <lu_zero> there aren't many solutions about it
[14:52:24] <lu_zero> either you know that beforehand or you need to write it somewhere
[14:52:48] <lu_zero> the filename comes to mind as ugly hack that works
[14:53:52] <mru> the flaw in pkgconfig is it assumes everything will use the same compiler
[14:54:05] <lu_zero> true
[14:56:29] <lu_zero> so if the compilers had a wider standard subset of options probably we would safe
[14:56:37] <lu_zero> would be
[14:56:51] <mru> that would be one solution
[14:57:19] <mru> but pkgconfig has many more issues
[15:36:10] <Honoome> mru: between libtool archives and pkg-config I still prefer pkg-config… the same goes for most $foo-config scripts that tend to make even more mistakes
[15:37:37] <mru> when the choices are death or syphilis...
[15:38:53] <lu_zero> mru: I consider it less tragic
[15:39:13] <av500> mru: syphilis, at least you had some fun before :)
[15:39:23] <lu_zero> the choices are between something completely broken and something that might not work always
[15:40:00] <Honoome> you know… you could fall back to autoconf m4 macros *cough* *cough* *cough*
[15:40:03] <Honoome> okay it's not funny, I know
[15:40:19] <mru> pkgconfig doesn't support cross-compiling, period
[15:40:36] <Honoome> that's correct
[15:40:41] <Honoome> do $foo-config support it? :P
[15:40:48] <mru> no
[15:41:07] <Honoome> btw I think I proposed a solution upstream a long time ago but they didn't bother listening to me :/
[15:41:09] <mru> ff check_lob does
[15:41:11] <mru> lib
[15:41:16] <Honoome> [as it happens a lot with fd.o]
[15:45:02] <lu_zero> mru: but that fails in some situations as well
[15:45:18] <lu_zero> sadly we haven't a perfect solution for everything (yet)
[15:45:56] <lu_zero> so pkg-config might be a nice and simple solution for problems such "guess the deps in my static lib"
[15:46:43] <lu_zero> but is quite annoying if you don't have the host-triplet-pkg-config machinery set
[15:46:48] <lu_zero> and you want to cross compile
[15:47:18] <lu_zero> (and that machinery feel overly complicated and more or less fixed just in their bzr tree)
[15:47:32] <av500> lu_zero: would the libs it depends on change depending on whether you cross?
[15:48:51] <lu_zero> av500: yes =P
[15:49:11] <lu_zero> we have most of that stuff done in gentoo cross wrappers
[15:49:44] <lu_zero> so you have one pkgconfig and host-triplet-wrappers for it that set the correct path where to search .pc files
[15:50:24] <lu_zero> autotools will privilege the wrapper over the native pkgconfig
[15:50:46] <lu_zero> so at least for them the thing works as expected
[15:51:02] <mru> the prefix approach would require me to have a separate copy of pkgconfig for every *system*, not just one per arch
[15:51:16] <lu_zero> a bash script that sets 2 variables
[15:51:27] <mru> and even that shouldn't be needed
[15:52:02] <lu_zero> you set CC and such when you cross compile already...
[15:52:22] <lu_zero> and you have already some shorthands to pick a prefix and use it for all the tools you are going to use
[15:52:30] <mru> I don't have one compiler per system
[15:52:33] <lu_zero> so it isn't _that_ bad
[15:52:50] <mru> one arm compiler is enough
[15:52:57] <av500> mru: pmi, but what is the solution?
[15:53:03] <lu_zero> neither I do ^^
[15:53:32] <mru> they could support --sysroot like gcc does
[15:53:51] <mru> or an env var with the same effect
[15:54:11] <mru> now the root is hardcoded into the binary
[15:54:16] <mru> twats
[15:54:22] <lu_zero> there is an env root for it
[15:54:23] <lu_zero> BUT
[15:54:35] <mru> it doesn't work
[15:54:40] <lu_zero> they messed up and didn't release a fix for it
[15:54:55] <mru> and it adds -L/usr
[15:54:58] <lu_zero> I should check if leio fixed it for good in gentoo at least
[15:55:04] <lu_zero> mru: got fixed
[15:55:08] <mru> and then linking fails
[15:55:25] <lu_zero> _it_got_fixed_in_their_bzr_
[15:55:26] <mru> not motivated to look at it
[15:55:48] <mru> I have a workaround I can live with
[15:56:12] <lu_zero> anyway I should check leio did it or I'll do myself eventually =P
[16:01:22] <lu_zero> pkg-config-.25 released....
[16:02:04] <peloverde> Is it worth adding a DCT-IV or should I just try to abuse the MDCT?
[16:05:22] <lu_zero> bbl
[16:05:35] <mru> whichever is faster
[16:06:35] <peloverde> DCT-IV could be faster but only after new ASM gets written
[16:07:06] <mru> how much faster?
[16:10:33] <mru> and what do you need it for?
[16:10:57] <peloverde> The SBR decoder analysis filterbank
[16:11:54] <mru> and how much faster?
[16:11:59] <mru> estimate
[16:12:15] <j0sh_> BBB: what would be preferable rather than returning on size==0? (in my first patch)
[16:12:42] <j0sh_> another ff_http* api function and an additional flag?
[16:13:18] <peloverde> It will save and O(N) pre-process step
[16:14:29] <BBB> j0sh_: no, I was just wondering if that was needed
[16:14:34] <BBB> j0sh_: what do you need that for?
[16:14:57] <BBB> does read(0) fail without that?
[16:14:58] <j0sh_> oh, it's to avoid a hang on the url_read for rtsp-http
[16:15:15] <j0sh_> because we don't get anything in the response until we send out the POST
[16:15:26] <j0sh_> but of course we can't send out the POST until the GET returns
[16:15:31] <BBB> ah, so you need to init the connection before sending the post?
[16:15:32] <j0sh_> and so forth
[16:15:34] <j0sh_> right
[16:15:37] <BBB> got it
[16:15:38] <BBB> ...
[16:15:40] <BBB> uhm...
[16:15:52] <peloverde> I suppose it's not that much, but generating a bunch of zeros to feed into the MDCT only to get refolded seems a bit silly
[16:16:24] <BBB> j0sh_: ok leave it in for now :)
[16:16:31] <mru> it does...
[16:16:42] <j0sh_> alright cool, new patch coming right up
[16:18:09] <BBB> cool, can probably apply
[16:18:12] <BBB> on to the next stage then
[16:19:29] <j0sh_> \oo/
[16:19:55] <BBB> you have two faces?
[16:20:10] <BBB> that's a little odd
[16:20:16] <BBB> ;)
[16:20:27] <BBB> did you talk to spyfeng btw?
[16:21:51] <j0sh_> ohh, that's a face. always thought it was a hand
[16:22:19] <j0sh_> not yet, i just woke up. i think he's probably still sleeping if he's in china
[16:22:30] <j0sh_> \n
[16:25:21] <mru> a
[16:27:02] <BBB> \oo/ looks more like glasses to me
[16:27:08] <BBB> well anyway
[16:27:10] <BBB> good work :)
[16:31:31] <Tjoppen> oh dear. I just figured out the ridiculously circuitous way pix_fmt is carried over to the rawvideo decoder via codec_tag, and why that doesn't work for mxf
[16:35:44] <mru> ouch
[16:36:00] <mru> codec_tag is best ignored
[16:39:26] <Honoome> I seriously hate C++ fanboys who can't even read…
[16:39:55] <mru> you mean all of them?
[16:41:16] <j0sh_> c++ was my first love
[16:41:26] <Honoome> mru: in particular those who can only read the first paragraph of a post and comment with shit on it…
[16:41:27] <j0sh_> but you know how first loves sometimes are. it's just a bitter memory now
[16:41:37] <Tjoppen> it turns out I have to leave bits_per_coded_sample along in mxfdec.c if it successfully detects the pixel format
[16:41:43] <Honoome> j0sh_: haha :) that's one quoteworthy line :D
[16:41:53] <Tjoppen> utils.c and rawdec.c transport it via codec_tag. that makes no sense at all
[17:03:07] <ramiro_> hyc: http://pastebin.com/BbJcwM74 ?
[17:04:47] <janneg> av500: the beast has landed
[17:04:53] <av500> cool
[17:05:09] <av500> there is a small power tool inside
[17:05:17] <av500> should be all that is needed to set up
[17:05:37] <janneg> suitcase is heavier than expected
[17:05:41] <av500> :)
[17:05:54] <av500> please do not use the wheels for long distance travel if possible
[17:06:02] <av500> I doubt they will survive
[17:06:11] <janneg> cool, I was going to ask which tools I should bring
[17:06:16] <av500> :)
[17:33:00] <Kano> hi, when will multichannel vorbis work (needed for webm)
[17:33:15] <Kano> encode works but playback not, getting similar errors like
[17:33:18] <Kano> https://roundup.ffmpeg.org/issue1969
[17:33:21] <elenril> when someone implements it?
[17:33:27] <elenril> send patches ;)
[17:33:31] <av500> Kano: patches welcome
[17:34:16] <Kano> well i cant patch this, the vorbis dev say ffmpeg devs should fix it ;)
[17:34:29] <lu_zero> janneg: should I bring more chocolate?
[17:34:46] <peloverde> Kano, you can always hire someone to
[17:35:04] <Kano> is it possible to force usuing libvorbis to decode?
[17:36:07] <kierank> does <video> support surround sound?
[17:36:15] <lu_zero> kierank: yes
[17:36:27] <lu_zero> iirc
[17:36:35] <kierank> is this de facto support or is it actually specified?
[17:36:45] <av500> <video> does not specify anthging
[17:36:47] <lu_zero> kierank: I could look the w3 spec
[17:37:04] <kierank> hmmm...so <video> does have something over flash ;)
[17:38:25] <lu_zero> The video  element is a media element  whose media data is ostensibly video data, possibly with associated audio data.
[17:38:42] <lu_zero> flash is video + canvas more or less
[17:38:52] <lu_zero> http://www.w3.org/TR/html5/video.html
[17:38:58] <kierank> thanks
[17:39:23] <kierank> surround sound needs to be standardised imo because of all the varying channel orders etc
[17:40:16] <av500> ...In addition to the above, the user agent may provide messages to the user (such as "buffering"...
[17:40:26] <av500> so, Real was part of the standardisation?
[17:41:40] <lu_zero> ...
[17:41:48] <lu_zero> it's http
[17:42:16] <lu_zero> so you'll need to buffer in order to get the extradata
[17:42:17] <peloverde> I feel like the real buffering problem was more an artifact of the era more than anything else
[17:42:38] <lu_zero> peloverde: think a mov with moov at the end
[17:42:48] <peloverde> qt-faststart
[17:43:05] <av500> lu_zero: why on earth would anybody even think of using such a thing for http video?
[17:43:10] <Kano> firefox 3.6 can play the ogg file inside the browser
[17:43:21] <lu_zero> or an ogg with loots of "useful" metadata in the packet 2
[17:43:23] <Kano> from the demo site
[17:43:29] <lu_zero> and you need packet 3
[17:43:50] <peloverde> I want a version of this with the firefox instead of the fish http://imagebin.ca/view/m8GFEJ.html
[17:43:54] <av500> lu_zero: well, buffering 3 packets is different from seeking to random position in the video...
[17:44:15] <av500> peloverde: pixels welcome
[17:44:16] <kierank> so, what's the best way of getting ffmpeg to handle dependent streams, for h.264 mvc
[17:45:27] <lu_zero> av500: metadata packet in ogg
[17:45:37] <lu_zero> guess what you can put there...
[17:45:45] <av500> lu_zero: what?
[17:45:47] <lu_zero> hints: poster image
[17:45:54] <lu_zero> author full bio
[17:46:01] <av500> and?
[17:46:03] <lu_zero> pink elephants dancing
[17:46:58] <lu_zero> so between playing guess-where-moov-is
[17:47:10] <lu_zero> and fetch something unbounded
[17:47:31] <lu_zero> you might have usage for a notice like "buffering"
[17:47:43] <lu_zero> even if I'd use "gone hunting, be back soon"
[17:47:56] <av500> lu_zero: of course, but you might also use a file format that minimizes hunting....
[17:48:52] <lu_zero> webm?
[17:48:55] <lu_zero> nut?
[17:49:02] <av500> asf!
[17:49:04] * av500 hides
[17:49:19] <lu_zero> mov with moov at the top?
[17:49:25] <av500> no
[17:49:27] <peloverde> ts?
[17:49:35] <lu_zero> mpegps!
[17:49:35] <av500> u cant stream live content with moov at the top
[17:49:45] <lu_zero> why not?
[17:49:48] <av500> raw video and audio over 2 urls?
[17:50:06] <kierank> flv as well
[17:50:42] <peloverde> how about ditch http?
[17:50:43] <lu_zero> Kano: I think that probably it's an extension xiph coded
[17:50:50] <lu_zero> rtsp?
[17:50:52] <peloverde> also does fragmented mp4 work for live streaming?
[17:51:00] <av500> it should
[17:51:05] <lu_zero> peloverde: apple wrote something about
[17:51:53] <lu_zero> anyway
[17:52:37] <lu_zero> mru: you'd like to try another troll beer?
[18:02:20] <BBB> yuvi: ok, now it uses dsp funcs and ff_emulated_edge_mc, it looks almost correct, I'll go track down the load of remaining bugs in my code... is intra bitexact already?
[18:05:15] <lu_zero> BBB: is something I can try as well?
[18:05:18] * lu_zero curious
[18:05:31] <BBB> lu_zero: check guthub/yuvi branch vp8
[18:05:35] <BBB> github*
[18:17:06] <Kano> could somebody port the m2ts time calculation from xbmc? there the time is correct. what vlc shows and other apps has nothing to to with the runtime
[18:18:18] <av500> patches welcome...
[18:29:32] <lu_zero> Kano: fill an issue pointing the exact data, that would make things easier
[18:36:13] <Kano> why is aac integrated marked as experimental
[18:36:41] <BBB> it's not as good as faac yet
[18:36:57] <janneg> Kano: try to encode a low bitrate stereo stream
[18:37:20] <Kano> at least it is better than vorbis ;)
[18:43:04] <lu_zero> Kano: eh eh
[18:45:20] <av500> can haz somebody add iphone voice call to ffmpeg?
[18:45:30] <av500> plz send me the codes
[18:48:02] <lu_zero> ?
[18:48:31] <av500> http://www.blogcdn.com/www.engadget.com/media/2010/06/apple-wwdc-2010-384-rm-eng.jpg
[18:49:08] <kierank> iAcronyms are soo cool
[18:49:38] <av500> iAgree
[18:53:38] <j0sh_> you take all those acronyms together, you have v/voip
[18:54:41] <kierank> suddenly those acronyms are super-important now that jobs has mentioned them...
[18:54:43] <j0sh_> can iphone's dsp encode h264 in realtime?
[18:54:52] <av500> it seems it can
[18:55:02] <av500> they talk about 720p30fps encoding
[18:55:28] <janneg> or they added another one
[18:55:37] <j0sh_> 720p? wow
[18:56:26] <j0sh_> you know what's funny though
[18:56:36] <av500> j0sh_: not really wow, that is state of the art
[18:56:38] <janneg> there are also several android phones anounced with 720 recording
[18:56:45] <av500> yup
[18:57:32] <j0sh_> i've always found h263 calls to be of better quality than 264
[18:57:46] <j0sh_> but maybe my phone's encder sucks, dunno
[18:58:15] <j0sh_> my 264 calls are blurry because of heavy deblocking
[18:59:28] <j0sh_> (http://www.viable.net/product/vpad)
[18:59:53] <av500> well, it will be interesting to see ip apple can bring video calls forward
[19:00:06] <j0sh_> i hope they can. i'm staking my future on it
[19:00:29] <av500> you bought stocks in quickMakeUp.co?
[19:00:33] <kierank> lol video calls, 2003's calling
[19:00:55] <j0sh_> heh no, seriously
[19:01:03] <j0sh_> im gonna do a startup in that space after gsoc
[19:01:06] <av500> lol url: http://www.bbc.co.uk/food/recipes/carrotsglazedwithcum_80467
[19:04:49] <j0sh_> twilio for video calls, what do you guys think?
[19:12:53] <Kano> bye
[19:36:02] <Yuvie> BBB: intra should be bitexact if the video uses the simple loop filter
[19:36:02] <Yuvie> which it doesn't unless you encoded it with ultra fast settings
[19:36:02] <Yuvie> or if you used enough bitrate to disable the loop filter, I guess
[20:00:47] <janneg> mru: is your mobile number still the same as last year? I plan to get there early with the beast.
[20:01:10] <mru> yep, still the same
[20:01:17] <mru> if anyone needs my number, whois has it
[20:02:10] <mru> I got an email from wired asking about the video wall
[20:03:23] <janneg> if there is enough space I can organize a car for the lcd later
[20:03:39] <mru> I don't know what vlc and xbmc plan to bring
[20:03:46] <mru> I assume at least one display each
[20:05:52] <av500> mru: wired, cool
[20:06:49] <mru> he wants to know what the idea behind it was
[20:06:57] <mru> and what the goal is
[20:07:02] <mru> sort of the same question...
[20:07:59] <janneg> I'll bring the posters from KotH, 8 port switch, network cables, power extensions, tape etc.
[20:08:12] <janneg> do we need anything else?
[20:08:52] <kierank> mru: which wired?
[20:08:56] <mru> .com
[20:09:16] <kierank> ah ok
[20:11:16] <av500> mru: tell him much of it comes from 5 19" lcds sitting unused in a spare room :)
[20:12:28] <av500> mru: www.vp7.de/videowall.avi
[20:12:38] <mru> what's that?
[20:12:47] <av500> the 1st videowall :)
[20:13:26] <av500> its not synced, but using the isp resizer for each unit to upscale its sector
[20:16:37] <mru> seen that before
[20:16:45] <av500> yeah
[20:19:19] <BBB> Yuvi: ok so it's not? :) thing is, testing for bitexactness of inter seems pointless if intra isn't, I mean, it uses framedata of intra...
[20:19:52] <Yuvie> BBB: just encode at a really high bitrate so that the loop filter has no effect, dunno how quickly I'll be able to fix the normal filter
[20:20:19] <BBB> it's ok, don't worry about it for now... it looks quite good so I might just go lazy and wait and do nothing ;)
[20:20:23] <BBB> this "works for me"
[20:24:25] <lu_zero> janneg: may I nag you about directions ?
[20:24:46] * lu_zero is reviewing the Berlin map
[20:24:49] <janneg> lu_zero: sure
[20:25:40] <mru> proceed down the street until you find an u-bahn or s-bahn station
[20:25:47] <mru> take the first train going in your direction
[20:26:27] <lu_zero> the map service was suggesting me picking a bus
[20:26:43] <janneg> from tegel?
[20:26:54] <mru> from tegel you take the bus
[20:27:00] <mru> there are a few different ones
[20:27:23] <mru> I take the first to arrive and get off when it stops at a rail station of some kind
[20:27:30] * mru doesn't like buses
[20:27:39] <lu_zero> janneg: yes
[20:27:54] <lu_zero> my target is Xantener Strasse 4
[20:28:09] <janneg> http://www.fahrinfo-berlin.de/Fahrinfo/bin/query.bin/en?ld=0.1& can be trusted
[20:28:39] <lu_zero> maps says that's I could even might walk for a bit south and I'll reach the point
[20:29:06] <lu_zero> but http://www.vbb-fahrinfo.de/hafas/query.exe/en?seqnr=1&ident=9o.0156101.1275942194&OK#focus
[20:29:10] <lu_zero> that's what I got
[20:29:53] <mru> try google maps for walking directions from southampton to edinburgh
[20:30:05] <mru> it starts by sending you on a ferry to france
[20:30:16] <mru> then back to england
[20:30:20] <lu_zero> uhmm
[20:30:20] <mru> then to france again
[20:30:28] <av500> it beats walking
[20:30:37] <mru> google's walking directions include ferries but not trains or buses
[20:30:39] <lu_zero> you did somethink to somebody at google, haven't you?
[20:32:38] <janneg> lu_zero: I would guess it's a 10km walk from tegel to xantener str
[20:34:47] <lu_zero> 7.2
[20:35:16] <janneg> so take the bus. either directly to "U Adenauerplatz" with the '109' or with the 'X9' to "U Jakob-Kaiser-Platz" and change the U7 to "U Adenauerplatz"
[20:36:01] <lu_zero> ok
[20:37:08] <lu_zero> we'll meet at around 2 at the same entrance of last year?
[20:38:12] <janneg> works for me, but we could also meet at the booth since we don't need special setup tickets
[20:38:38] <lu_zero> how's working for the tickets?
[20:39:02] <mru> we collect them on arrival
[20:39:25] <mru> you should all be on the list
[20:39:29] <lu_zero> ok
[20:55:09] <KotH> mru: i didnt order any tickets for mplayer+ogp
[20:55:28] <KotH> mru: if anyone asks, they should know everything already as i wrote some mails about that
[20:55:53] <mru> ok
[21:35:47] <CIA-92> ffmpeg: cehoyos * r23515 /trunk/libavcodec/libvpxenc.c:
[21:35:47] <CIA-92> ffmpeg: Use new macro AV_BASE64_SIZE.
[21:35:47] <CIA-92> ffmpeg: Patch by James Zern, jzern google com
[21:45:56] <mchinen> looks like i missed the linuxtag chat :(
[21:50:47] <janneg> mchinen: just join us tomorrow and chat from face to face
[21:52:58] <CIA-92> ffmpeg: alexc * r23516 /trunk/libavcodec/ (aacsbr.c sbr.h): Rewrite the SBR decoder QMF analysis filter on top of the IMDCT instead of the RDFT.
[21:53:17] <mchinen> 2pm at the messgelaende?
[21:54:58] <janneg> 2-3 pm messegelaende / messe sued / halle 7
[21:56:28] <mru> peloverde: did you decide against the new dct?
[21:56:50] <mchinen> ok see ya guys there.  will you guys be obvious?
[21:56:52] <peloverde> mru, I had forgotten that IMDCT is close enough for my purposes
[21:57:13] <mru> so you're not shuffling about zeros after all?
[21:57:23] <peloverde> nope
[21:57:32] <peloverde> imdct_half is nice
[21:58:03] <mru> that it is
[21:58:17] <mru> so now it's using loads of asm too, yay
[21:59:42] <janneg> mchinen: we are the ones handling 6 19" displays
[22:00:41] <mchinen> k thx
[22:13:35] <peloverde> Now to do the synthesis fb
[22:29:36] <BBB> Dark_Shikari: ok, help needed again :)
[22:30:38] <BBB> Dark_Shikari: let's say it works a little, but there's some visual artifacts (especially in high-motion parts). how do I know figure out what code is faulty, except for comparing it value-for-value with the ref decoder? is there an easier way?
[22:31:30] <Dark_Shikari> BBB: find the first MB that differs
[22:31:32] <Dark_Shikari> find what type that MB is
[22:31:35] <Dark_Shikari> add printfs
[22:32:04] <BBB> ok, so compare value-per-value with the ref decoder basically?
[22:32:15] <Dark_Shikari> sorta, but targeted.
[22:32:21] <BBB> :)
[22:38:47] <mru> the nature of the artefacts can give clues too
[22:41:11] <Yuvie> I like diffing the decoded output and blowing up errors, to get stuff like http://i.imgur.com/O8bAO.png
[22:41:54] <Yuvie> http://pastie.org/995713 <- my lazy program to do that on rawvideo
[22:50:44] <mru> aac sbr needs asm
[22:51:18] <astrange> i use http://gitorious.com/~astrange/ffmpeg/ffmpeg-mt/blobs/master/mt-work/yuvcmp.c + printfs
[22:51:25] <mru> it's spending 78% time in the top 3 functions on arm
[22:51:38] <mru> 45% in ff_sbr_apply
[22:51:56] <mru> most of that is trivially simdable
[22:52:10] <Dark_Shikari> how slow is sbr on arm?
[22:52:13] <Dark_Shikari> like, what X realtime?
[22:54:28] <mru> 5.6x faster than realtime on cathedral
[22:55:20] <mru> ~60x faster on non-sbr
[22:55:33] <Dark_Shikari> lol
[22:55:39] <Dark_Shikari> 60x
[22:55:42] <mru> that's gcc sucking at floating-point
[22:55:44] <Dark_Shikari> "high efficiency" indeed
[22:56:17] <mru> I estimate I can make those 3 funcs taking 80% time 10x faster
[22:56:49] <CIA-92> ffmpeg: alexc * r23517 /trunk/libavcodec/aacsbr.c: 10l: Fix types and now unused variables after last commit.
[22:56:50] <peloverde> mru, I'm rewriting the synth bank now
[22:56:53] <astrange> llvm's arm support should be mature by now, it might work better
[22:56:57] <peloverde> Analysis should be ready for SIMD
[22:57:02] <astrange> depending on which part of the compiler matters for this
[22:57:32] <mru> this is very hard to do right on A8
[22:57:37] <mru> for a compiler
[22:57:57] <mru> I can probably write the asm in less time than it takes to build the compiler
[22:58:25] <mru> at least if you include the time required to figure out what's broken in the build system _this time_
[22:58:35] <peloverde> DS, High efficiency means high compression ratio, like High Efficiency laundry detergent
[22:58:57] <Dark_Shikari> I know ;)
[22:59:54] <peloverde> There is actually an SBR variant with an all real filterbank called LP SBR
[23:00:14] <mru> low probability?
[23:01:11] <peloverde> Low Power
[23:01:44] <mru> power as in signal power or compute power?
[23:01:54] <peloverde> compute power
[23:03:02] <mru> anyhow, the 3 time-eating functions are ff_sbr_apply, autocorrelate, and sbr_qmf_synthesis
[23:03:15] <mru> the first has loads of stuff inlined
[23:04:18] <peloverde> ff_sbr_apply is the top level function, hence the ff_ symbol
[23:04:20] <peloverde> autocorrelate is a bunch of unaligned dot products
[23:04:30] <peloverde> sbr_qmf_synthesis is the synthesis filterbank
[23:05:16] <mru> I kinda guess that...
[23:11:13] <mru> btw, have you noticed there's a long-running thread on vlc-devel about seeking in ogg
[23:11:27] <mru> evidently something about it isn't quite as simple as some would have you believe..
[23:12:51] <peloverde> That's odd, the conventionally wisdom is that Monty-god-of-all-media successfully refuted who-the-hell-is-mru?
[23:14:04] <Honoome> peloverde: if you intend media as communication outlets, you have two Montys to choose from
[23:17:36] <ohsix> are you talking about the thread thats on and on about indenting and stuff?
[23:19:27] <iive> ohsix: i thought that everybody agreed on strict K&R .
[23:19:41] <ohsix> iive: no idea with vlc, tbh
[23:19:51] <BBB> Yuvie: thanks for the hint
[23:19:55] <iive> oh...
[23:20:16] <ohsix> i'm just looking for a "long running thread on vlc-devel about seeking in ogg", and the only one i can find is like 80% "whitespace & comments wrong" stuff from the patch
[23:21:13] <ohsix> oh and he's fighting with is editor :] http://mailman.videolan.org/pipermail/vlc-devel/2010-June/075190.html
[23:22:42] <mru> I didn't read the thread
[23:23:03] <ohsix> oic
[23:23:04] <mru> there's never anything interesting in ogg threads
[23:23:21] <ohsix> well its a "how do i send a patch or work emacs" thread, huhuhuhu
[23:23:50] <mru> but even the fact that a patch large enough to have whitespace trouble is _possible_ should be a hint
[23:24:28] <ohsix> he appropriated code from another project; conventions differed, most of the posts are about removing artifacts of that project in the comments
[23:26:31] * iive wonders what that another project is...
[23:26:48] <ohsix> "lives", iive; click on the url i posted its got the goods
[23:27:22] <ohsix> he's sending mail with html attachments too, so how about that
[23:27:49] <peloverde> If I have two methods of doing something and all else is equal except one reverses a vector the other multiplies every other item by -1 what's preferable?
[23:28:09] <ohsix> its apparently a GSOC project and the patch is like 50kb
[23:29:06] <ohsix> 1454 insertions, 122 deletions
[23:30:01] <bcoudurier> anybody familiar with ipad development ?
[23:30:09] <Dark_Shikari> a bit
[23:30:18] <bcoudurier> I hate http live streaming
[23:30:30] <BBB> I can do iphone development
[23:30:34] <BBB> ipad should be the same
[23:31:01] <bcoudurier> did you work with http ls ?
[23:33:08] <kierank> 2000 line patch just to seek in ogg :/
[23:33:32] <ohsix> "and misc"
[23:33:46] <ohsix> the way its titled it seems to indicate its just to do theora in ogg properly
[23:34:06] <ohsix> instead of just landing in the middle of the bitstream and guessing at the length of the title
[23:34:08] <Yuvie> yeah, you need other stuff to do ogm correctly
[23:34:21] <Yuvie> dirac might work too depending on how it's structured
[23:43:17] <Honoome> gha… and I'm again struggling to find a way to just know much generically much time does it take to run a tokenizer…
[23:44:20] <Honoome> lu_zero found a new feature in Ragel we weren't using before (scanners) but after porting my previous manual code to that, the size of the code (at -O0) increased waaaay too much
[23:47:25] <BBB> bcoudurier: no, sorry :(
[23:47:30] * BBB goes home
[23:47:58] <BBB> Yuvie: you should submit the vp8 decoder for inclusion in svn, it's not too bad ;)
[23:48:35] <Dark_Shikari> lol
[23:51:01] <janneg> could be marked as experimental


More information about the FFmpeg-devel-irc mailing list