[FFmpeg-devel-irc] IRC log for 2010-02-11

irc at mansr.com irc at mansr.com
Fri Feb 12 01:00:16 CET 2010

[00:00:34] <DonDiego> how come nobody can be arsed to fix warnings nowadays?
[00:12:50] <Honoome> DonDiego: I should really try to find time to come back for that :P
[00:31:58] <DonDiego> gnite
[01:10:46] <peloverde> Is there a good way to do a fractional frequency shift on the FFT of an all real signal?
[01:11:45] <peloverde> There is definitely symmetry in the FFT so it should be exploitable but I don't know how
[01:20:08] <ramiro> mru: do you happen to know why sometimes when "make -j2" or "make -j3" is run, some .o files get deleted at the end? For example I just got this: rm doc/ffserver.pod ffmpeg.o doc/ffplay.pod doc/ffmpeg.pod
[01:20:39] <ramiro> that ffmpeg.o doesn't seem to belong there.
[02:33:54] <Compn> [21:41] <bjsnider> thbe entire ubuntu multimedia system from gstreamer to vlc to mplayer all uses external shared ffmpeg
[03:38:24] <BBB> merbzt, I've started writing wmavoice specs for the wiki, comments would be appreciated if you have time
[04:20:50] <justlooking_>  /msg NickServ identify
[06:44:13] <KotH> bonjour!
[06:44:21] <kshishkov> gruss dich
[06:46:47] <elenril> http://news.slashdot.org/story/10/02/10/2347257/Subversive-Groups-Must-Now-Register-In-South-Carolina wait, what?
[06:47:02] <kshishkov> subversive, not SVN
[06:47:26] <elenril> also morning
[06:47:40] <KotH> elenril: old! :)
[06:51:02] <pJok> mornings :)
[06:51:09] <superdump> morning
[06:51:19] <kshishkov> morgnar
[06:52:09] <KotH> ah.. people wake up! :=)
[06:53:27] <kshishkov> okay, now what?
[06:57:30] * kshishkov prepares for sleep mode again
[06:57:31] <KotH> let's start a revolution and overthrow a goverment
[06:57:58] <kshishkov> nah, we had that several times already
[06:59:03] * elenril overthrows KotH 
[06:59:09] <kshishkov> there were half a dozen of different governments in Ukraine during 1917-1919 for example and there were also revolutions in 1991 and 2004
[06:59:53] * KotH is no goverment
[07:00:17] <KotH> lets overthorw the UN!
[07:00:29] <KotH> they are standing in our way to world domination anyways!
[07:00:30] <kshishkov> KotH: how else can we call root then?
[07:00:47] <KotH> kshishkov: benelovent dictator for lifetime? ;-)
[07:02:03] <kshishkov> KotH: that is the definition of government, yes
[07:02:29] <KotH> no, i dont govern you, i just dictate ;)
[07:02:47] <kshishkov> nobody governs me
[07:03:10] <kshishkov> here you quickly learn the differences between country and homeland
[07:03:23] * elenril kicks intel drivers for randomly disabling lvds yet again
[07:04:16] <KotH> lol
[07:04:24] <KotH> kshishkov: get your ass moving and move out of ua
[07:04:56] <kshishkov> KotH: I try, but there are issues with receiving side :(
[07:22:10] <superdump> kshishkov: what issues are there on the receiving side?
[07:22:57] <kshishkov> superdump: well, in order to move to another country one need some permit
[07:23:07] <kshishkov> in your case citizenship is enough
[07:23:24] <kshishkov> in my case I need explicit permit
[07:23:43] * pJok makes the ukraine a member of the EU
[07:23:50] <kshishkov> had I such thing, I'd move without any hesitation
[07:24:02] <superdump> how can you get a permit?
[07:24:17] * kshishkov tells pJok that even omnipotent deity can't do that
[07:24:50] <kshishkov> superdump: that's a tricky question
[07:24:59] <pJok> looking at which countries that got into the eu, ukraine has a fair chance
[07:25:02] <superdump> but i guess it's the one you want answered
[07:25:32] <kshishkov> pJok: nope, it does not have any chances because it's screwed its chance already
[07:27:19] <pJok> is the ukraine in a worse economic condition than greece?
[07:27:47] <kshishkov> superdump: here, for example - http://swedenabroad.com/Page____48659.aspx
[07:27:59] <kshishkov> probably it is
[07:28:24] <kshishkov> where can I compare stats?
[07:30:18] <superdump> kshishkov: does that apply to _anyone_ visiting sweden regardless of where they come from?
[07:30:36] <DonDiego> libavcodec/indeo5.c:676: warning: implicit declaration of function ‘ivi_calc_band_checksum’
[07:30:45] <DonDiego> kshishkov: please don't ignore such warnings
[07:31:00] <DonDiego> they're trivial enough to fix..
[07:31:14] <kshishkov> DonDiego: ok, will fix
[07:31:27] <DonDiego> but please don't just ignore warnings
[07:31:43] <pJok> kshishkov, wiki has a lot of different comparisons between countries
[07:31:47] <DonDiego> the number of warnings has skyrocketed lately
[07:32:00] <kshishkov> superdump: probably not, there are certain advantages for EU members and even more advantages for .no/.dk/.fi
[07:32:26] * pJok can become a swedish citizen in april
[07:32:41] <pJok> not that i have any need for that
[07:33:43] <pJok> thats because in april i've lived two years in sweden
[07:33:50] <superdump> kshishkov: why is there this special case: International passport (for ukrainian citizens) ?
[07:34:06] <superdump> do ukranian citizens not have to provide the copies of all pages?
[07:34:27] <kshishkov> pJok: .ua inflation >12%, .gr inflation < 3%
[07:34:54] <superdump> aaah
[07:34:56] <superdump> i see
[07:35:02] <superdump> it was the ukraine specific page
[07:35:02] <pJok> well, greece is in big trouble at the moment
[07:35:04] <superdump> :)
[07:35:06] <kshishkov> superdump: ask DonDiego for more details, they also have that system with internal and international passports
[07:35:39] <pJok> germany has that as well
[07:35:47] <pJok> personenausweiss and passport
[07:36:14] <kshishkov> I like Swedish way though
[07:36:45] <pJok> in sweden, as in denmark, you only have an international passport
[07:37:15] <pJok> the id you either get from the government or what ever bank you are stuffing your moneey into
[07:37:25] <kshishkov> or post
[07:37:54] <superdump> kshishkov: http://swedenabroad.com/Page____63492.aspx
[07:38:01] <superdump> looks like it's a little easier for me
[07:38:28] <superdump> i move here, within 3 months of moving here i tell them i'm here and i have a right to be here
[07:38:49] <superdump> then i sort out the swedish ID number within my first year though the sooner the better
[07:38:50] <kshishkov> told you
[07:38:51] <superdump> awesome
[07:38:59] <kshishkov> that's because you're EU citizen
[07:39:03] <superdump> i was a bit concerned when looking at the work permit thing
[07:39:21] <superdump> for that your company has to advertise the job here and blah
[07:39:35] <superdump> it's actually easier to just move here and sort out the job stuff afterwards
[07:40:04] <kshishkov> yes, but in my case there is no easy way to move at all
[07:40:21] <superdump> aaah
[07:40:32] <superdump> advertised the job vacancy in sweden and the EU
[07:41:47] <superdump> kshishkov: working in sweden as an EU/EEA citizen - http://swedenabroad.com/Page____101519.aspx
[07:42:39] <superdump> kshishkov: have you applied for a permit?
[07:43:42] <kshishkov> superdump: how? Benjamin has not adopted me yet, no company has sent me an offer and my self-unemployment plan is not good either
[07:44:01] <superdump> pJok: i thought ID cards only allowed one to travel around the EU and they weren't valid outside, where you would actually need a passport
[07:44:37] <superdump> merbzt: is it legal to have two wives (well, a wife and a husband) in sweden?
[07:44:39] <superdump> :)
[07:45:30] <superdump> kshishkov: hrm, i dunno. but your programming skills should make you hirable
[07:46:09] <kshishkov> superdump: I've heard it was only legal in Stockholm in sixties
[07:46:50] <kshishkov> superdump: but my status of citizen of third-world country blocks hiring me
[07:48:04] <superdump> why?
[07:48:23] <superdump> some open source friendly company with a brain could see that you're very capable from your work on ffmpeg
[07:48:33] <kshishkov> because of aforementioned procedure of hiring
[07:49:29] <superdump> thanks for the link to that page by the way
[07:49:37] <superdump> i had been looking for information on how to move here
[07:50:35] <DonDiego> ffmpeg branch in sweden?
[07:51:01] <kshishkov> I saw all people who form that branch
[07:55:29] <CIA-17> ffmpeg: kostya * r21752 /trunk/libavcodec/indeo5.c:
[07:55:29] <CIA-17> ffmpeg: Move band checksum verifying into preprocessor condition, so compiler won't
[07:55:29] <CIA-17> ffmpeg: complain about missing function prototype.
[07:59:48] <kshishkov> DonDiego: okay, one warning less
[07:59:53] <kshishkov> N to go ;)
[08:00:17] <CIA-17> ffmpeg: kostya * r21753 /trunk/libavcodec/indeo5.c: Move 'chksum' declaration to the only block where that variable is used
[08:05:00] <astrange> the h264 warnings are easy to remove, some entirely unused variables and two missing &s in (*blah)[x] stuff
[08:08:14] <DonDiego> it's just a shame that michael doesn't give a hoot about warnings
[08:08:32] <kshishkov> GCC does not help either
[08:09:02] <DonDiego> all of these are correct warnings
[08:11:11] <astrange> i introduced two of them myself and forgot to send the patch fixing them
[08:11:24] <astrange> guess i'll do it friday
[08:11:42] <DonDiego> just commit..
[08:12:03] <DonDiego> what's so controversial about removing unused vars?
[08:12:21] <astrange> well, nothing
[08:15:57] <siretart> god morgon
[09:17:51] <av500> hi guys
[09:18:23] <merbzt> hello
[09:18:24] <kshishkov> hi guy
[09:18:43] <andoma> hej
[09:19:05] <av500> btw, there is another ffmpeg groupie? http://www.youtube.com/watch?v=iIalNEW-LQ8&feature=fvw
[09:20:28] <kshishkov> we are official at least
[09:26:00] <iive> what? uau forked ffmpeg too?
[09:27:09] <elenril> http://kotaku.com/5469239/hey-korean-kids-lets-learn-leetspeak-and-internet-slang << lol
[09:27:31] <elenril> iive: i think he has a mirror on repo.or.cz with some -mt related patches
[09:27:54] <elenril> +merged libswscale
[10:00:11] <merbzt> av500: awesome video :)
[10:05:24] <av500> merbzt: what is also fun is to search ebay.com for ffmpeg :)
[10:07:26] <kshishkov> what "FFmpeg r20359, one owner, slightly used"?
[10:58:47] <mru> I have a suggestion
[10:58:55] <mru> let's find out what's broken on fate and fix it
[11:03:03] <kshishkov> FATE should print that
[11:03:13] <mru> fate doesn't say why
[11:07:14] <kshishkov> look at the tests, something ate away last frames
[11:07:35] <mru> some tests are failing only on some configs
[11:09:05] <kshishkov> still it has something to do with time handling changes in ffmpeg.c
[11:09:23] <mru> so find out what's correct and fix what's not
[11:09:28] <mru> we can't keep it like this
[11:10:26] <kshishkov> well, 307 passed seems normal
[11:10:43] <kshishkov> 303 passed means VP6 segfaulted
[11:11:10] <mru> and 304?
[11:11:29] <kshishkov> I meant 304
[11:11:37] <mru> well, 303 then?
[11:11:49] <kshishkov> for 303 it's full regtest failing too
[11:12:02] <mru> ah, those are 304 + old failures
[11:12:08] <kshishkov> in mp2 decoder
[11:14:10] <kshishkov> it's funny that on BeagleBoard full regtest fails because of mpeg4 asp encoder
[11:14:27] <mru> with rvct?
[11:14:53] <kshishkov> yes
[11:14:59] <mru> compiler bug
[11:15:02] <mru> already fixed
[11:18:03] <av500> kshishkov: wrt ebay, mostly webhosting with ffmpeg preinstalled...
[11:22:48] <kshishkov> av500: yay, so now it's like ImageMagick
[11:23:32] <merbzt> ___gb___: have you done any work with the CrystalHD rstuff ?
[11:27:50] <jez9999> kshishkov: you have ffmpeg checking access?
[11:27:52] <jez9999> *checkin
[11:28:05] <merbzt> all ops should have
[11:28:09] <jez9999> ok
[11:28:13] <jez9999> could you check in a minor bugfix?
[11:28:24] <jez9999> i've submitted the patch1.diff in https://roundup.ffmpeg.org/roundup/ffmpeg/issue1740
[11:28:24] <kshishkov> theoretically yes
[11:29:09] <merbzt> no note that it is ok'd
[11:29:21] <merbzt> didn't look at the actual patch
[11:29:22] <jez9999> note from whom
[11:29:32] <jez9999> it's so small it should be easy to quickly review
[11:29:47] <merbzt> from the maintainer that the patch is ok
[11:29:53] <mru> FFmagick
[11:30:56] <kshishkov> mru: well, certain Compn always wanted FFmpeg to be an ubiquitious image converter as well
[11:31:33] <jez9999> and the maintainer is...?
[11:32:05] <kshishkov> Michael Niedermayer
[11:32:15] <kshishkov> (at least I think so)
[11:32:22] <jez9999> then why was BBB saying he'd "accept" the patch?
[11:32:22] <merbzt> if not in the MAINTAINERS it is MN
[11:32:28] <kshishkov> aka default dev/maintainer
[11:33:29] <jez9999> [15:52] <@BBB> jez9999, I said, add a ?opt=val to the sdp file
[11:33:29] <jez9999> [15:53] <@BBB> ./ffplay file.sdp?opt=val
[11:33:54] <merbzt> jez9999: I guess because he knows that the patch is correct
[11:33:54] <jez9999> [15:57] <jez9999> the sdp handler doesn't get invoked, the file protocol handler fails
[11:33:54] <jez9999> [15:57] <@BBB> hm, I guess the file doesn't exist... crap, this is how rtsp fixes these issues
[11:33:55] <jez9999> [15:58] <@BBB> ok, then go fix that bug also then as part of this bug ;)
[11:33:58] <merbzt> I don't
[11:34:34] <jez9999> well you have his word for it now :-)
[11:34:37] <jez9999> dont you trust him?
[11:34:45] <kshishkov> on the contrary
[11:34:52] <kshishkov> we trust him to commit this patch
[11:35:15] <kshishkov> in the case of something, he'll be able to explain what is it for and why it was done so
[11:35:24] <jez9999> gah
[11:35:31] <merbzt> :)
[11:35:44] <jez9999> that means i have to wait for him to get out of dozyland in the damn USA
[11:35:56] * kshishkov does not know a thing about protocols, especially RT*P
[11:36:00] <merbzt> yes
[11:36:18] <jez9999> they should just adjust their clocks to be in line with ours and work at night
[11:36:39] <jez9999> where does he live, NY?
[11:36:42] <merbzt> suggest that to the appropriate place
[11:37:13] <kshishkov> jez9999: nah, be thankful _they_ did not make you adjust your clock
[11:37:26] <kshishkov> Airstrip One
[11:37:56] <__gb__> merbzt, not yet, just compiled the driver on my system for now :)
[11:38:09] <jez9999> is it NY?
[11:38:59] <kshishkov> jez9999: no, it's how Great Britain is called in British Government Development Plan (aka "1984")
[11:39:11] <jez9999> where he lives i mean
[11:39:21] <merbzt> __gb__: there was a post on ffmpeg-devel about it, might be possible to fund the work
[11:39:59] <kshishkov> yes, your guess is correct
[11:40:05] <superdump> donation offer for crystal hd support in ffmpeg on the ML
[11:40:08] * superdump looks at GB
[11:40:11] <jez9999> probably 2-3 hours then..... at least....
[11:40:14] <superdump> __gb__: ^
[11:40:20] * jez9999 twiddles thumbs
[11:42:31] <kshishkov> superdump: quite an ambiguous two-letter abbreviation, isn't it?
[11:42:35] <merbzt> __gb__: well I've only just installed it in my machine
[11:43:40] <merbzt> lspci finds the card
[11:43:48] <superdump> may as well get the funding for th ework
[11:43:50] <superdump> the work*
[11:44:09] <superdump> as it's being offered
[11:44:17] <superdump> is it just a hardware decoder card?
[11:44:20] <kshishkov> well, "donation" usually hints at smaller sums
[11:44:22] <superdump> why is there so much noise about it?
[11:44:23] <merbzt> yes
[11:44:37] <merbzt> because they released drivers for it
[11:44:41] <merbzt> and it works
[11:44:51] <kshishkov> and it's rather opensource
[11:45:31] <elenril> is there any sane reason why they didn't use vdpau or vaapi?
[11:45:53] <kshishkov> is there any sane reason to use it?
[11:46:25] <elenril> to not have yet another acceleration api?
[11:46:41] * elenril heard that vdpau is NotTerrible
[11:46:55] <CIA-17> ffmpeg: kostya * r21754 /trunk/libavutil/sha.c:
[11:46:55] <CIA-17> ffmpeg: Make SHA digest function write digest value with AV_WN32 instead of assuming
[11:46:55] <CIA-17> ffmpeg: that output may be written as uint32_t since output buffer may not be aligned
[11:46:55] <CIA-17> ffmpeg: (and it's silly to force alignment on it) and it does not work in that case
[11:46:55] <CIA-17> ffmpeg: properly on some architectures.
[11:47:04] <__gb__> vdpau or vaapi are decode+display APIs, not really suitable to crystalhd, which needs the whole bitstream
[11:47:16] <superdump> is crystalhd hella cheap or something?
[11:47:34] <merbzt> 21$ was what I payed for my card
[11:47:38] <__gb__> crystalhd is not quite opensource, the whole stuff is hidden in a big 1 or 2 MB firmware
[11:47:39] <superdump> :)
[11:47:45] <superdump> that's the answer then
[11:47:48] <kshishkov> elenril: also both those APIs look like inhouse API released for public
[11:48:02] <superdump> $21 is quite a bit cheaper than a vaapi/vdpau supporting card i guess
[11:48:04] <__gb__> yes, this is the same api for windows, linux and macos x
[11:48:36] <merbzt> __gb__: would it be possible to wrap it under vaapi ?
[11:49:00] <__gb__> this could be hard because crystalhd needs the whole bitstream, vaapi or vdpau work at picture level
[11:49:07] <merbzt> ok
[11:49:26] <__gb__> e.g. you'd have to reconstruct sps and pps headers, which can be hard without much changes to vaapi or vdpau
[11:49:50] <__gb__> but I not thought much about it yet
[11:49:51] <merbzt> then I guess the best way is to hook it up into the ffmpeg api
[11:50:02] <__gb__> yes, but there might also be a problem
[11:50:18] <__gb__> iirc, scott told me that crystalhd requires a few encoded frames prior to emitting the first decoded frame
[11:50:42] <__gb__> but I probably misunderstood the thing
[11:50:52] * __gb__ is going to cook and eat -- see you later :)
[11:53:49] <CIA-17> ffmpeg: siretart * r21755 /branches/0.5/ (libavcodec/avcodec.h Changelog):
[11:53:49] <CIA-17> ffmpeg: reverting objected hunks from previous commit
[11:53:49] <CIA-17> ffmpeg: as discussed with diego on irc, the spurious newline deletion and the
[11:53:49] <CIA-17> ffmpeg: LIBAVCODEC_VERSION_MINOR bump are being reverted based on comments on
[11:53:49] <CIA-17> ffmpeg: ffmpeg-cvslog by ramiro, uoti and michael.
[11:53:50] <CIA-17> ffmpeg: See http://comments.gmane.org/gmane.comp.video.ffmpeg.cvs/28112 for the
[11:53:51] <CIA-17> ffmpeg: full context.
[11:54:09] <ramiro> DonDiego: ping. did you read the gsm thread?
[12:13:38] <superdump> merbzt: they're going to put out an expresscard version so people can munge it into their laptops and netbooks :D
[12:13:54] <superdump> i wonder what the power usage is on one of those broadcom doodads
[12:14:36] <merbzt> they are doing pciexpress cards also
[12:15:16] <superdump> yup
[12:19:34] * kshishkov would like low-power system solely for 1280x760 H.264 playback though
[12:20:00] <twnqx> too small!
[12:20:41] <twnqx> though i'd like a PoE dongle with a HDMI connector on the other end, and sp/dif in the middle.
[12:21:14] <mru> kshishkov: omap3
[12:24:53] <superdump> does anyone here watch much 1080p content yet?
[12:25:26] <DonDiego> ramiro: no
[12:25:34] <Dark_Shikari> I do
[12:25:44] <twnqx> superdump: i dop
[12:25:46] <twnqx> do*
[12:25:50] * mru too
[12:25:51] <Dark_Shikari> thanks to hdbits
[12:26:29] <twnqx> do they give out invites again, btw?
[12:27:10] <kshishkov> mru: BB seems to be slow for that :(
[12:27:28] <ramiro> DonDiego: please do. it's currently inconsistent.
[12:27:28] <Dark_Shikari> twnqx: still no
[12:27:34] <twnqx> we should make michael optimize ffmpeg until HE can watch 1080p content
[12:27:37] <mru> you need av500's 800MHz omap3 and some fancy codecs
[12:27:39] <mru> it can be done
[12:28:19] <twnqx> someone told me of an anime rip where the end exceeds 100mbit bitrate in the end...
[12:28:29] <twnqx> foir one scene
[12:29:07] <Dark_Shikari> not surprising
[12:29:38] <twnqx> 100mbit is realistic with (numerically) low CRF and high movements, i guess?
[12:29:40] <superdump> big buck bunny plays fine using one core on my core 2 vpro 1.83GHz
[12:29:46] <Dark_Shikari> BBB is _EASY_
[12:29:50] <Dark_Shikari> I encoded that at 1mbps
[12:29:51] <superdump> i thought as much
[12:29:52] <Dark_Shikari> looks great
[12:29:54] <superdump> ha
[12:29:56] <superdump> :)
[12:30:00] <Dark_Shikari> with weightp, even moreso
[12:30:03] <mru> BBB has mostly static backgrounds
[12:30:04] <superdump> it's very low motion really
[12:30:05] <Dark_Shikari> because at 1mbps the only noticably awful scene was the fade
[12:30:37] <mru> the zoom at the start and the pan where the rat barely avoids various traps are the hardest parts
[12:31:01] <Dark_Shikari> btw
[12:31:06] <Dark_Shikari> here's my two awesome "quality" images
[12:31:07] <Dark_Shikari> http://i48.tinypic.com/2rnhf06.png
[12:31:11] <Dark_Shikari> dvd source
[12:31:16] <Dark_Shikari> due to the "requires 100mbit" phenomenon
[12:31:24] * Dark_Shikari uploads similar example from blu-ray
[12:31:30] <Dark_Shikari> http://i46.tinypic.com/2nu0nqv.png
[12:32:12] <superdump> urrrgh
[12:32:31] <twnqx> haruhi!
[12:32:47] <twnqx> with horrible artefacts >_>
[12:32:52] <superdump> i don't have the hard drive space on this laptop to get 1080p stuff
[12:33:05] <superdump> nor a 1080p cable subscription or whatever
[12:33:15] <twnqx> no bittorent?
[12:33:28] <twnqx> 1080p bluray rips are in the range of 8-11G
[12:33:28] <superdump> nor the bandwidth to be bothered to wait for 1080p stuff to download
[12:33:30] <twnqx> (normally)
[12:33:54] <superdump> plus i don't have a 1080p screen here yet :)
[12:34:19] <superdump> i wondered if maybe there were some bravia + ps3 bundles to save me getting a BD player
[12:34:21] <mru> superdump: get job, get money, get screen
[12:34:29] <mru> don't buy a bravia
[12:34:32] <mru> waste of money
[12:34:33] <superdump> have job, getting money, will get screen shortly
[12:34:38] <superdump> i have a samsung at home
[12:34:43] <mru> samsung or lg are as good and half the price
[12:34:49] <superdump> hehe
[12:34:54] <superdump> which should cover a ps3 anyway
[12:34:58] <Dark_Shikari> WDTV is $80
[12:35:00] <Dark_Shikari> it plays 1080p
[12:35:01] <mru> ps3 is good
[12:35:02] <superdump> i did want an led backlit one, but....
[12:35:02] <Dark_Shikari> you do not need anything else
[12:35:19] <superdump> Dark_Shikari: can you plug in a BD drive? :D
[12:35:28] <Dark_Shikari> why would you need a BD drive?
[12:35:54] <superdump> in case i bought some blu ray discs
[12:35:56] <Dark_Shikari> >implying that people actually buy blu-ray movies
[12:36:37] <mru> Dark_Shikari: where did you think the rips came from?
[12:38:01] <superdump> Dark_Shikari: does the wdtv have any fans or anything or is it passive?
[12:38:09] <Dark_Shikari> mru: yes, but _one_ person buys it
[12:38:10] <Dark_Shikari> at most
[12:38:12] <KotH> Dark_Shikari: what's teh source of these two images?
[12:38:12] <Dark_Shikari> superdump: dunno
[12:38:15] <Dark_Shikari> it's a hardware player
[12:38:18] <Dark_Shikari> KotH: 1st is a Haruhi DVD
[12:38:22] <Dark_Shikari> not sure about the blu-ray one
[12:38:31] <superdump> it would be nice to have no fans, no moving parts and no electrical whining either
[12:38:42] <twnqx> <Dark_Shikari> >implying that people actually buy blu-ray movies <-- i do
[12:38:45] <KotH> Dark_Shikari: the second image is from utawarerumono
[12:39:02] <superdump> but i guess using an SSD as mass storage isn't advisable yet
[12:39:02] <janneg> superdump: it has a harddisk
[12:39:07] <KotH> Dark_Shikari: eh.. you mean that was professionaly produced stuff??
[12:39:08] <superdump> janneg: right
[12:39:24] <superdump> what kind of ui does it have? what software?
[12:39:33] <Dark_Shikari> KotH: yes
[12:39:37] <superdump> have people hacked it?
[12:39:45] <superdump> can you stick a bittorrent client on it? :)
[12:39:57] <Dark_Shikari> the first one simply ran out of bitrate, combined with a bad MPEG-2 encoder
[12:40:03] <Dark_Shikari> because that scene will exceed ~9mbps
[12:40:06] <Dark_Shikari> it hit QP31
[12:40:14] <Dark_Shikari> The second one is inexcusable, but also MPEG-2
[12:40:27] <Dark_Shikari> R2 DVDs were better for Haruhi, that scene was just lowpassed instead of blocked to hell
[12:42:06] <KotH> Dark_Shikari: i've to have a look at the fansub... but i dont remember seeing anything that bad ...
[12:42:32] <Dark_Shikari> it probably wasn't blu-ray based
[12:43:28] <av500> superdump: wdtv no fans afaik
[12:44:56] <peloverde> good morning
[12:45:00] <av500> gm
[12:47:08] <superdump> so, basically you need a wdtv and some kind of NAS which runs a bittorrent client and has a usb port on it so that all its storage can be accessed as mass storage
[12:47:29] <superdump> the only thing then is whether the wdtv presents the stuff on the drive in some useful way
[12:47:40] <superdump> i suspect xbmc or so might be a nicer interface
[12:47:46] * twnqx uses a atom/ion htpc
[12:48:10] <kshishkov> superdump: theoretically gigabit network performance should be enough
[12:48:34] <superdump> indeed, having the whiney storage elsewhere is an option
[12:49:02] * kshishkov uses NAS made from SheevaPlug+2.5" USB HDD
[12:50:38] <av500> superdump: wdtv live has eth, so u need a san somewhere...
[12:51:46] <Dark_Shikari> gigabit is plenty
[12:51:49] <Dark_Shikari> the decoder can't even handle that much
[12:52:04] <av500> h264 UHP?
[12:52:41] <superdump> http://www.trustedreviews.com/multimedia/review/2009/10/29/Western-Digital-WDTV-Live-HD-Media-Player/p1 <--- the video here recommends the asus hd pr1 or something like that as it's just as good but quite a bit cheaper
[12:53:50] <Dark_Shikari> cheaper than ~80-100$?
[12:53:51] <av500> superdump: but my guy "in the know" tells me that these days the xstreamer is a better choise than wdtv life: http://www.xtreamer.net/
[13:44:45] <lu_zero> yawn
[13:46:54] <KotH> bon giorno lu_zero
[13:48:09] <lu_zero> hi KotH
[13:55:06] <neo01124> asd
[14:09:35] <CIA-17> ffmpeg: kostya * r21756 /trunk/libavutil/sha.c: Simplify expression as suggested by M?ns Rullg?rd
[14:28:54] <jez9999> just to check (and i doubt this), are there functions already in ffmpeg to 1) split a query string from a URL, and b) break up the query string's key/value's into some kind of array?
[14:29:23] <kshishkov> 1) url_parse()
[14:29:34] <kshishkov> 2) probably none
[14:29:54] <KotH> tr /12/ab/
[14:31:30] <jez9999> url_parse looks PHPish
[14:31:59] <kshishkov> wanna pick a fight?
[14:32:04] <jez9999> where is it defined?
[14:32:50] <KotH> kshishkov: i'd take that as a yes ;)
[14:32:59] <kshishkov> libavformat/avformat.h - url_split()
[14:36:06] <jez9999> kshishkov: i dont see that giving you the query string back anywhere
[14:36:47] <kshishkov> probably in path
[14:37:31] <jez9999> oh well.  that's doing a lot more work than i need anyway
[14:38:17] <kshishkov> it's so trivial nobody ever bothered to write a support for it in URL splitter
[14:39:15] <jez9999> yeah
[14:39:20] <jez9999> i wrote something to do it
[14:41:21] <KotH> y
[15:00:37] <jez9999> BBB: hi
[15:08:49] <BBB> hi jez9999
[15:09:07] <jez9999> so from yesterday:
[15:09:25] <jez9999> [15:52] <@BBB> jez9999, I said, add a ?opt=val to the sdp file
[15:09:26] <jez9999> [15:53] <@BBB> ./ffplay file.sdp?opt=val
[15:09:26] <jez9999> [15:57] <jez9999> the sdp handler doesn't get invoked, the file protocol handler fails
[15:09:26] <jez9999> [15:57] <@BBB> hm, I guess the file doesn't exist... crap, this is how rtsp fixes these issues
[15:09:26] <jez9999> [15:58] <@BBB> ok, then go fix that bug also then as part of this bug ;)
[15:09:42] <jez9999> i have a patch to fix that bug, can you commit it?
[15:10:15] <BBB> submit it to the mailinglist
[15:10:20] <BBB> or to the issue tracker
[15:10:25] <BBB> let's see if michael's ok with it
[15:10:28] <BBB> I have no strong opinion
[15:10:38] <BBB> he might say "no, go work on codec/format-specific options"
[15:11:37] <jez9999> it's already in the issue tracker.
[15:14:27] <jez9999> https://roundup.ffmpeg.org/roundup/ffmpeg/issue1740
[15:14:33] <jez9999> patch2.diff
[15:14:35] <jez9999> so what now?
[15:14:50] <jai> drink beer?
[15:15:01] <andoma> yes!
[15:15:06] <KotH> no!
[15:15:41] <andoma> ok, in a few hours ..
[15:16:16] <jai> andoma: what time is it there?
[15:16:55] <andoma> 16:16
[15:17:16] <jai> ah, k
[15:17:31] <BBB> jez9999, oh, I was sick yesterday so I missed it - will check
[15:34:16] <jez9999> BBB: thanx
[16:04:02] <BBB> Vitor1001, wmavoice patch ok with you?
[16:04:18] <Vitor1001> have to go now, but not yet :(
[16:05:15] <Vitor1001> I have just one more comment, but I really think it would be nice to get another pair of eyeballs to look at three funcs
[16:05:23] <Vitor1001> aw_parse_coords()
[16:05:34] <Vitor1001> aw_pulse_set{1,2}()
[16:05:56] <Vitor1001> I cannot see any more how to improve them, but maybe I just read them too many times
[16:06:07] <BBB> Reimar also commented a lot on them
[16:06:12] <BBB> I think they're a lot prettier already
[16:06:16] <Vitor1001> and when one do that one miss some things...
[16:06:22] <BBB> right :)
[16:06:31] <BBB> http://wiki.multimedia.cx/index.php?title=Windows_Media_Audio_Voice#Frames
[16:06:35] <BBB> trying to document also
[16:06:48] <Vitor1001> yeah, but it really would be nice if someone looked at it for the first tine after all the simplifications
[16:06:53] <Vitor1001> \o/
[16:07:09] <BBB> you shoudl document sipro
[16:07:11] <Vitor1001> My comment about it is in aw_parse_coords()
[16:07:15] <BBB> I'd like to link to it for relevant stuff
[16:08:09] <Vitor1001> I think (s->aw_n_pulses[idx] == 0),  (first_idx[n] > 0) and (s->aw_first_pulse_off[idx] == NO_OFFSET) are equivalent
[16:08:24] <Vitor1001> IMHO, using everywhere just the first one is clearer if it is the case
[16:08:37] <BBB> yeah
[16:08:39] <BBB> ok
[16:08:45] <Vitor1001> I'm lazy about documentation
[16:08:51] <Vitor1001> writing text in general
[16:08:56] <BBB> also, the first loop (where I set nmin/nmax) and the second (where I use them) can probably be merged
[16:08:58] <BBB> I'll try that
[16:08:59] <Vitor1001> ideally I should document also TwinVQ :(
[16:09:20] <Vitor1001> Have to go now, be back in ~ 30 min
[16:09:22] <kshishkov> ideally you should document every codec in existence
[16:21:08] <jez9999> BBB: you taken a look yet?
[16:23:52] <CIA-17> ffmpeg: siretart * r21757 /branches/ (0.5 0.5/libavcodec/snow.c):
[16:23:52] <CIA-17> ffmpeg: Fix crash when max_ref_frames was out of range.
[16:23:52] <CIA-17> ffmpeg: This might have been exploitable.
[16:23:52] <CIA-17> ffmpeg: Fixes first crash of issue840.
[16:23:52] <CIA-17> ffmpeg: backport r18388 by michael
[16:25:25] <BBB> jez9999, working on other things right now, I will
[16:25:31] <jez9999> k
[16:35:46] <Vitor1001> I'm back
[16:36:12] <kshishkov> what about new thoughts on WMASpeech?
[16:36:26] * kshishkov want FFmpeg to decode all formats ASAP
[16:36:51] <Vitor1001> kshishkov: ?
[16:36:54] <kierank> need moar samples
[16:36:59] <jai> instant world domination eh?
[16:36:59] <kierank> ;)
[16:37:30] <kshishkov> Vitor1001: I meant WMS review
[16:37:50] <Vitor1001> Read my conversation with BBB 20 min ago
[16:38:01] <kshishkov> I read
[16:38:02] <Vitor1001> I was just looking for someone else to review three functions
[16:38:33] <Vitor1001> Do you volunteer to review them after BBB applies my suggestions?
[16:38:35] <BBB> I simplified aw_parse_coords()
[16:38:40] <Vitor1001> Nice
[16:38:41] <BBB> it has one loop now
[16:38:42] <BBB> instead of two
[16:38:46] <BBB> I think you'll like it
[16:38:48] <BBB> let me pastebin it
[16:38:51] <kshishkov> Vitor1001: ok, I am
[16:39:28] <Vitor1001> BBB: did you suceed in removing the first_off[] array?
[16:39:54] <BBB> http://ffmpeg.pastebin.com/m9dd635d
[16:39:55] <BBB> Vitor1001, yes
[16:39:56] <Vitor1001> And removing the use of NO_OFFSET in  aw_parse_coords()
[16:40:00] <BBB> no
[16:40:06] <BBB> NO_OFFSET should stay, at least for now
[16:40:10] <Vitor1001> Why?
[16:40:15] <superdump> KotH, mru : i'm going to be running some progressive download tests on mphq samples for all the files for a few container formats, is that ok?
[16:40:16] <BBB> no time yet :]
[16:40:20] <BBB> will work on it
[16:40:31] <superdump> i won't need to read the entire file in any case, so i shouldn't be using too much bandwidth
[16:40:41] <BBB> (I know, you want me to use aw_n_pulses[] and leave aw_first_pulse_off uninitialized
[16:40:52] <superdump> but just to let you know in case you see someone hammering the http access to samples.mphq
[16:41:02] <BBB> nmin/nmax also removed again
[16:41:05] <BBB> should've known :)
[16:41:09] <BBB> anyway
[16:41:09] <KotH> superdump: ip?
[16:41:13] <Vitor1001> BBB: What I suggested was instead of " if (first_idx[n] > 0)"
[16:41:15] <superdump> errrm
[16:41:23] <Vitor1001> doing "if (s->aw_n_pulses[idx])"
[16:41:25] <KotH> superdump: ip range is ok too :)
[16:41:29] <BBB> Vitor1001, that's not equivalent
[16:41:30] <Vitor1001> and removing the first_idx[] array
[16:41:50] <BBB> don't forget that this is positioning of a repeating pulse
[16:41:59] <BBB> first_idx simply tells us where the first pulse was:
[16:42:03] <BBB> before the start of block1
[16:42:04] <Vitor1001> I see
[16:42:05] <BBB> in block1
[16:42:07] <BBB> or in block2?
[16:42:14] <BBB> I can make that a variable also
[16:42:25] <Vitor1001> wouldn't (s->aw_first_pulse_off[n] != NO_OFFSET) do it them?
[16:42:36] <BBB> no
[16:42:48] <BBB> I could store the offset = start_offset[bits] value
[16:42:50] <BBB> that would do
[16:42:58] <BBB> if <0, then first_idx>0 for both
[16:43:14] <Vitor1001> ugh, the negative coefficients again :p
[16:43:14] <BBB> if >= 0 but < MAX_FRAMESIZE /2, only for [1]
[16:43:18] <BBB> else for neither
[16:43:27] <BBB> hey I didn't design this codec :)
[16:43:29] <Vitor1001> I understand...
[16:43:31] <Vitor1001> I know ;)
[16:43:38] <BBB> let me try getting rid of first_idx[]
[16:43:39] <BBB> sec
[16:43:55] <Vitor1001> But that's why I want someone else to look at it... I'm starting to misread this func...
[16:45:39] <Vitor1001> BTW, kshishkov, thanks :)
[16:47:50] <BBB> same uri has next version
[16:47:59] <BBB> I caled the variable first_off
[16:48:03] <BBB> as opposed to first_idx[]
[16:48:12] <BBB> maybe the loop at the end can be simplified also now
[16:48:42] <BBB> since it makes no sense to continue the loop for n=0 if it was false for n=1
[16:48:56] <BBB> so I should do for (n = 1; n >= 0; n--) if .. else break;
[16:49:19] <BBB> but I could just as well manually unroll it
[16:49:21] <BBB> hmm...
[16:49:49] <BBB> anywyay, let me know if this is better
[16:49:52] <kshishkov> hmm, why can't you replace first for() loop with division?
[16:50:10] <Vitor1001> Nice
[16:50:12] <BBB> hmm... I don't need n anymore
[16:50:14] <BBB> let me remove it
[16:50:25] <BBB> kshishkov: because I thought I needed n
[16:51:03] <Vitor1001> Just do not try to optimize it if it makes less readable
[16:51:12] <Vitor1001> doesn't looks like a very speed critical func.
[16:51:20] <kshishkov> have you corrected line 39 to offset += pitch[idx] ?
[16:51:39] <BBB> done
[16:51:59] <BBB> division you mean offset %= pitch[0];?
[16:52:08] <kshishkov> yes, like that
[16:53:04] <Vitor1001> MS engineers designed this part of wmav during a trip in hell :p
[16:53:36] <kshishkov> see XKCD for "Ballmer's peak" definition
[16:54:33] <Vitor1001> I know it ;)
[16:56:11] <BBB> the %= doesn't work for negative numbers
[16:56:15] <kshishkov> also it looks like whole loopn on lines 31-40 can be boiled into directly setting pulse offsets and s->aw_n_pulses by dividing MAX_FRAMESIZE/2 by pitch
[16:57:26] <BBB> kshishkov, how do I know aw_first_pulse_off[]?
[16:58:17] <kshishkov> the only thing you need for it is offset
[16:58:57] <kshishkov> for the zeroeth pulse it's first offset >= 0, right?
[16:59:15] <kshishkov> for the first pulse it's first offset >= MAX_FRAMESIZE /2
[16:59:55] <kshishkov> which can be derived from initial_offset and number of times pitch[0] is used
[17:00:16] <kshishkov> which will also be the number for s->aw_n_pulses[0]
[17:00:27] <BBB> Vitor1001, ok, NO_OFFSET removed
[17:00:48] <kshishkov> determining that number is just one integer division
[17:01:09] <kshishkov> like (MAX_FRAMESIZE/2 + pitch[0] - 1 - first_offset)/pitch[0]
[17:01:24] <kierank> is twolame better than the ffmpeg mp2 encoder?
[17:01:47] <kshishkov> almost everything is
[17:01:53] <jai> most encoders are
[17:01:57] <jai> yep
[17:02:09] <jai> +audio
[17:02:17] <KotH> why doesnt ffmpeg have a mp1l3 encoder already?
[17:02:17] <kshishkov> I've heard our AC3 encoder is not bad
[17:02:32] <kshishkov> KotH: some political decision, I think
[17:02:57] <KotH> kshishkov: agreement with frauenhofer?
[17:03:03] <jai> KotH: we could write a simple one which outputs a valid bitstream
[17:03:28] <jai> but unless the psy stuff is working....
[17:03:30] <kshishkov> KotH: no, but personally I'd like to see MP3 buried. It's a very awful format
[17:03:43] <peloverde> I agree
[17:04:24] <kshishkov> and having data not bound to frames is maybe the worst thing ever
[17:05:46] <BBB> http://ffmpeg.pastebin.com/m6e294f05
[17:05:55] <BBB> is what I have now
[17:06:08] <BBB> the %= didn't work for the first loop, so tell me what it should be instead :)
[17:06:36] <mru> %, ouch
[17:07:24] <BBB> hey, he suggested it, not me :)
[17:07:27] * BBB ducks
[17:07:32] <Vitor1001> BBB: Nice about the NO_OFFSET! \o/
[17:07:52] <Vitor1001> mru: % is faster than doing a loop for evaluating it.
[17:08:12] <kshishkov> still, the second loop boils down to:
[17:08:15] <mru> only if you must do the calculation in the first place
[17:09:07] <kshishkov> set first_pulse_off[0]; set s->aw_n_pulses[0]; set s->aw_first_pulse_off[1]; set s->aw_n_pulses[1];
[17:10:11] <kshishkov> first can be done out of loop, second can be calculated by one division, third may be calculated from the second, tthe last can be calculated by simple division again
[17:11:09] <kshishkov> theoretically  you can replace first loop with if(start_offset < 0) offset = (start_offset % pitch[0]) + pitch[0];
[17:11:58] <kshishkov> well, it has some flaws but demonstrates my idea
[17:14:14] <superdump> KotH: do you know if http://samples.mplayerhq.hu/allsamples.txt is up-to-date?
[17:15:14] <BBB> kshishkov, flaws?
[17:15:26] <BBB> kshishkov, you're talking to someone that didn't know the restrict keyword until two days ago
[17:15:46] * kshishkov does not know about that thing even now
[17:16:20] <kshishkov> well, it misses "else" clause for one thing
[17:16:51] <Compn> superdump : its not up to date, but i can update it now if you want
[17:16:54] <BBB> is it going to be faster?
[17:17:07] <superdump> Compn: yes please
[17:17:09] <kshishkov> yes, and more branch predictable too
[17:17:10] <BBB> I mean, a simple while loop that executes (mostly) 0-1 times can't be that much slower than a "%"
[17:17:27] * BBB feels a wristslap coming
[17:17:28] <kshishkov> ah, in that case leave it as is
[17:17:30] <KotH> superdump: nope, it's not
[17:17:37] <KotH> superdump: 2009-11-11
[17:17:41] <mru> BBB: depends on how % is implemented
[17:17:57] <BBB> mru: uhm, by getting edx after idiv?
[17:18:06] * BBB feels another wristslap coming
[17:18:09] <Compn> superdump : refresh it
[17:18:15] <mru> you need to stop thinking in x86
[17:18:26] <BBB> it's the only asm I know :(
[17:18:29] <mru> and even there, the compiler might use a loop
[17:18:32] <KotH> superdump: now it's up to date :)
[17:18:41] <mru> is either operand a constant?
[17:18:46] <superdump> Compn: thanks :)
[17:19:14] <Compn> KotH : is there any specific command i should use to generate that file btw? :P
[17:19:15] <kshishkov> BBB: for the first loop it does not worth the salt, but you have still to replace the second one
[17:19:22] <KotH> Compn: no idea
[17:19:32] * Compn just using find | grep -v md5sum > allsamples.txt
[17:19:34] <KotH> Compn: i dont even know what for it is there :)
[17:19:41] <Compn> easy searching for files
[17:19:54] <Compn> like say you want to find all acelp samples
[17:20:02] <Compn> some are in A-codecs, some maybe in /mov/ or /real/ etc
[17:20:15] <BBB> kshishkov, ok, will try
[17:20:16] <KotH> there should be a database [tm]
[17:20:28] <Compn> then there are some other non-sorted dirs like A-codecs/suite or fate samples
[17:20:31] <jai> and xml!
[17:20:40] <BBB> so it's then (MAX_FRAMESIZE / 2 - offset) / pitch[0] for n_pulses?
[17:20:43] <KotH> Compn: ack
[17:20:54] * KotH gtg
[17:21:09] <kshishkov> BBB: probably + pitch[0] - 1
[17:21:40] <kshishkov> so offset will go over MAX_FRAMESIZE/2
[17:21:47] <kshishkov> as it's supposed
[17:22:38] <BBB> kshishkov, that's for offset, yes
[17:22:44] <BBB> but for n_pulses?
[17:22:51] <BBB> let me try
[17:22:56] * BBB no good at any of this
[17:22:58] <kshishkov> for n_pulses as well
[17:23:07] <kshishkov> but yes, I'd also try it
[17:26:46] <BBB> wow, it went correct the first time :)
[17:26:57] <BBB> completely loopless except for that thingy at the end
[17:28:12] <kshishkov> well, it could also be simplified in that way but it's shaving pigs
[17:29:38] <superdump> KotH: i find 2085 files i want to test
[17:29:42] <BBB> not worth it because (again) the loop will run 0-1 times, usually
[17:29:47] <superdump> does that file list symlinked files too...?
[17:29:56] <superdump> or just actual files
[17:29:56] <BBB> the middle loop that we just did runs 4-5 times on average
[17:29:59] <superdump> Compn: ^
[17:30:02] <kshishkov> BBB: that's why I said it's shaving pigs
[17:30:29] <kshishkov> superdump: probably it does, you can compare checksums though
[17:30:33] <BBB> http://ffmpeg.pastebin.com/m21e26d60
[17:30:37] <BBB> that's what I have now
[17:30:44] <superdump> bleh
[17:31:04] <BBB> Vitor1001, is the bottom of that OK with you?
[17:31:19] <BBB> Vitor1001, I'm unrolling it for the compiler b/c then I can inline the second if in the first
[17:31:28] <BBB> Vitor1001, apparently you never know with gcc
[17:31:39] <BBB> anyway, this function is about 10LOC less than the original
[17:32:12] <BBB> from 59 in my v1 patch to 45 now
[17:32:20] <kshishkov> BBB: looks more or less ok to me
[17:32:49] <BBB> want to do set2() also?
[17:32:52] <BBB> let me c/p it
[17:33:09] <kshishkov> maybe remove hash inside comment?
[17:33:17] <BBB> http://ffmpeg.pastebin.com/m671c3e4d
[17:33:31] <BBB> oh yeah, the hash is doxy leftovers, doxy links with the hash
[17:33:33] <BBB> but it's not doxy
[17:33:34] <BBB> so ok
[17:34:58] <Vitor1001> BBB: If it ok to kshishkov, I'm fine too.
[17:35:10] <BBB> it's a lot nicer now
[17:35:28] <kshishkov> Reimar would ask to use memset(use_mask + 5 instead of memset(&use_mask[5]
[17:35:32] <Vitor1001> BBB: I don't know if you need to unroll the loop, the code is not very speed critical
[17:35:45] <kshishkov> and sizeof(use_mask[0])
[17:35:47] <BBB> well I already did
[17:36:06] <BBB> kshishkov, he did, he didn't mind when I said I preferred this way
[17:36:07] <Vitor1001> BBB: If it is not less readable, fine for me too
[17:36:11] <BBB> ok
[17:36:44] <Vitor1001> But nice job you two, aw_parse_coords() is really neat now :D
[17:36:53] <BBB> sizeof(use_mask[0]) done
[17:37:11] <kshishkov> I'll look at the rest later, in 20-30 mins
[17:37:12] <kshishkov> bbl
[17:37:18] <BBB> sure, thanks!
[17:41:40] * Kovensky just saw http://hardwarebug.org/2010/02/10/1080p-video-on-beagle/ on reader
[17:41:50] * Kovensky recognizes nobody, probably because he doesn't know how anybody looks like
[17:42:21] <BBB> is there a reason to cache first-off? or is the compiler smart enough to know that start_off[bits] doesn't change given that all are local variables?
[17:43:30] <BBB> let's compare
[17:46:30] <BBB> aha, it's smaller
[17:46:32] <BBB> perfect
[17:53:52] <mru> Kovensky: I'm at the far right
[17:54:09] <mru> peloverde is at the other end
[17:54:12] <mru> the rest you can interpolate
[17:56:48] <mru> hmm... getopt(..., 'r:o:f:l:...')
[18:04:43] <kshishkov> BBB: in aw_pulse_set2 you can replace horrible idx adjust ment by simple loop checking on nonzero mask and idx = (mask_num << 4) + 0xF - av_log2(use_mask[mask_num])
[18:04:54] <kshishkov> (lines 50-56)
[18:08:02] <kshishkov> also if fcb->pitch_lag allows, it may be better to split idx into two variables - for lower 4 bits and high bits
[18:08:54] <kshishkov> two places it's used as one value are lines 49 and 61
[18:10:00] <BBB> kshishkov, it used to be that
[18:10:03] <BBB> kshishkov, this seemed faster
[18:10:09] <BBB> (although slightly more code)
[18:10:30] <BBB> reimar said that the loop code itself would require more registers than an unlooped version in total
[18:10:37] <BBB> which I tend to agree with
[18:10:41] <kshishkov> then make constants on lines 50-54 as hex values
[18:10:48] <kshishkov> it's clearer that way
[18:10:57] <BBB> 16*x+15?
[18:11:03] <BBB> (where x = use_mask idx)
[18:11:06] <BBB> ok
[18:11:07] <kshishkov> yes
[18:11:42] <Kovensky> mru: which right
[18:11:59] <mru> observer
[18:11:59] <kshishkov> Kovensky: photo right
[18:12:11] <Kovensky> ic
[18:12:16] <BBB> http://ffmpeg.pastebin.com/m58c7f2e4
[18:12:17] <mru> peloverde is the one with the beard
[18:12:25] <Kovensky> and ffmpeg shirt
[18:12:56] <BBB> unless you want me to write out 16 as 16 * 1 and . as 16 * 0
[18:12:57] <BBB> :)
[18:13:04] <BBB> (which I thought was sort of overkill)
[18:13:34] <kshishkov> BBB: btw, idx_sh and idx_mask meanings are kinda swapped
[18:14:11] <kshishkov> and I'd accept even simple 0x0F, 0x1F, 0x2F, 0x3F, 0x4F ;)
[18:14:33] <BBB> I think reimar & vitor disliked me using hes values
[18:14:37] <BBB> well anyway
[18:14:43] <BBB> I guess it doesn't matter so much
[18:15:02] * BBB will use hex then
[18:15:03] <BBB> smaller
[18:15:07] <kshishkov> still, I think you've got swapped names for idx_sh and idx_mask
[18:15:49] <BBB> I think the var names just suck
[18:16:07] <BBB> idx_sh = shifted index
[18:16:18] <kshishkov> still, I'd expect idx_sh used for shifts and idx_mask for mask index, not vice versa
[18:16:22] <BBB> idx_mask is silliest name ever
[18:17:42] <BBB> those names are just as bad
[18:17:45] <BBB> I agree though
[18:17:46] <BBB> bad names
[18:17:49] <BBB> let me think :)
[18:18:33] <BBB> idx_mask could be first_off
[18:18:35] <kshishkov> even simple 'mask' and 'shift' would be better
[18:18:36] <BBB> or no
[18:18:46] <BBB> mask should be off or so
[18:18:56] <BBB> and idx_sh just mask_idx
[18:19:01] <BBB> use_mask_idx
[18:19:02] <BBB> or so
[18:19:17] <BBB> I'll use a pointer into use_mask instead
[18:21:08] <kshishkov> the last two lines seems also the task for the Superhero who knows modular arithmetics
[18:21:59] <BBB> I was about to ask
[18:22:22] <BBB> probably (MAX_FRAMESIZE / 2 + pitch - 1 + offset) / pitch?
[18:22:26] <BBB> or something along those lines
[18:22:51] <kshishkov> pitch_lag - (FRAMESIZE/2 - start) % pitch_lag is more likely
[18:24:13] <kshishkov> n = (MAX_FRAMESIZE / 2 - start_off) % pitch_lag; if(n) n = pitch_lag - n;
[18:26:50] <BBB> n = (fcb->pitch_lag - 1 + MAX_FRAMESIZE / 2 - start_off) / fcb->pitch_lag;?
[18:27:09] <kshishkov> no, you don't need to divide here
[18:27:19] <BBB> you're doing a modulo, same thing
[18:27:20] <BBB> no?
[18:27:40] <kshishkov> you want to know a residue over HALF_MAX_FRAMESIZE
[18:27:54] <BBB> yes
[18:27:58] <Vitor1001> BBB: I dislike hex when not doing bit-based tricks
[18:28:29] <Vitor1001> for masks I have nothing against it.
[18:28:38] <kshishkov> Vitor1001: for traking compound variable whose low 4 bits have separate meaning it's fine too
[18:28:42] <BBB> Vitor1001, it's not a mask where kshishkov pointed it out
[18:28:59] <BBB> but it works fine here
[18:30:54] <BBB> kshishkov, your trick works too
[18:30:56] <BBB> and is shorter
[18:30:59] <BBB> I think I'll keep it
[18:31:08] <BBB>     n = (MAX_FRAMESIZE / 2 - start_off) % fcb->pitch_lag;
[18:31:08] <BBB>     s->aw_next_pulse_off_cache = n ? fcb->pitch_lag - n : 0;
[18:31:12] <kshishkov> and only one modulo instead of division and multiplication
[18:31:20] <BBB> trure
[18:32:50] <kshishkov> well, just throw a comment about 'idx' structure and that it's faster and it's fine with me
[18:35:04] <BBB> that its index in the 80-bit use_mask bit-array and thus (given that use_msk is 16-bit[5]) idx>>4 is the uint16_t index and the remainder (&15) is the bit index within?
[18:35:07] <BBB> or something else?
[18:35:47] <kshishkov> yes
[18:36:22] <kshishkov> saves some seconds on deciphering it and some minutes on answering the question why it was done so
[18:39:10] <BBB> got it
[18:45:36] <BBB> shall I resubmit this to the mailinglist? or is there more?
[18:45:49] <BBB> (you could also review the patch on the mailinglist to get the complete picture)
[18:46:05] <kshishkov> aw_get_pulses1() I suppose
[18:46:17] <kshishkov> the rest was fine I heard
[18:46:51] <BBB> aw_pulse_set1()?
[18:46:52] <kshishkov> and after that you can commit it directly I think
[18:47:32] <BBB> http://ffmpeg.pastebin.com/m6d4a52b6
[18:47:43] <kshishkov> yes, aw_pulse_set1()
[18:49:29] <peloverde> Do we have a knowlegebase of FFT or other transform tricks?
[18:49:47] <kshishkov> BBB: looks ok
[18:50:09] <kshishkov> peloverde: Wikipedia seems to have rather good theoretical articales on FFT
[18:50:23] <kshishkov> for the rest we have FFmpeg source code
[18:50:50] <peloverde> Wikipedia is actually pretty basic
[18:51:02] <mru> there are these things called books too...
[18:51:09] <BBB> oops
[18:51:50] <peloverde> I'm looking for fancy things like an FFT of an all real signal translated half a bin in the frequency domain
[18:51:57] <kshishkov> peloverde: I've looked at FFT there this morning, seems much nicer than usual coverage to me
[18:52:13] <kshishkov> BBB: that's a bad sign. What's happened?
[18:52:53] <kshishkov> peloverde: RDFT?
[18:52:55] <peloverde> kshishkov: There is nothing on the wikipedia page that isn't in my junior level text books
[18:53:05] <peloverde> shifted half a bin?
[18:53:35] <BBB> kshishkov, uhm, I tried to close a tab in my browser and then somehow focus was still on xchat
[18:53:57] <BBB> typical case of pebkac
[18:54:13] <kshishkov> peloverde: well, take generic formula, use zeroes where appropriate and get your new SBR transform
[18:58:40] <kshishkov> I suspect you can simply perform N/2 transform with post-twiddle
[19:01:10] <kshishkov> peloverde: wikipedia says you can do it as two-dimensional FFT - 2x(N/2), transpose and perform (N/2)x2 tranform
[19:01:59] <peloverde> kshishkov: The half bin translation
[19:02:03] <peloverde> I know how an RDFT works
[19:02:05] <BBB> kshishkov, Vitor1001: can you guys ok the patch? :)
[19:02:07] <peloverde> I wrote rdft.c
[19:02:11] <BBB> that way reimar will do stuff also
[19:02:13] <kshishkov> it's either whole N/2 transform may be skipped entirely or you can al least use 2-point FFT on (0; x) as special case
[19:02:35] <kshishkov> BBB: yes, fine
[19:02:55] <peloverde> The half bin translation is done by multiplying by a complex exponential in the time domain
[19:03:07] <peloverde> making the resulting signal no longer real
[19:03:22] <kshishkov> you can discard imaginary part
[19:04:08] * kshishkov wonders how FAAD(2) could get it
[19:05:30] <peloverde> I think I can still solve it like a normal RDFT only instead of transposing I have to do some adding
[19:05:51] <peloverde> I should still be able to demultiplex in the FFT domain because there is still symmetry
[19:29:45] <BBB> kshishkov, so if I have your code:
[19:29:45] <BBB>     n = (MAX_FRAMESIZE / 2 - start_off) % fcb->pitch_lag;
[19:29:45] <BBB>     s->aw_next_pulse_off_cache = n ? fcb->pitch_lag - n : 0;
[19:30:23] <Dark_Shikari> Yuvi: very very nice
[19:30:26] <BBB> would it be faster if I do something silly like s->aw.. = (fcb_pitch_lag * MAX_FRAMESIZE / 2 - MAX_FRAMESIZE / 2 + start_off) % fcb->pitch_lag;?
[19:30:39] <BBB> i.e. is the conditional something that slows it down considerably?
[19:30:53] <kshishkov> benchmark it!
[19:31:13] <BBB> time thinks that any variant I try takes +/- 0.4 sec to decode my 4 test files
[19:31:16] <BBB> it's not very accurate :)
[19:33:01] <kshishkov> hmm
[19:33:03] <BBB> I might need a benchmarking guide or so
[19:33:31] <kshishkov> look at START_TIMER/STOP_TIMER in FFmpeg code
[19:33:35] <astrange> run 5x in a row, take the 'user' time and then take the minimum
[19:33:38] <astrange> for large changes
[19:33:44] <astrange> otherwise use START/STOP timer
[19:33:48] <BBB> ok
[19:33:59] * BBB goes figure stuff out again
[19:34:05] <astrange> that's not absolutely accurate (it prints 600 dezicycles even if you do nothing) but relatively it's ok
[19:40:24] <Vitor1001> peloverde: The stuff I explained to you didn't work?
[19:41:58] <Vitor1001> BBB: ok'ed on list.
[19:42:00] <BBB> 1185 dezicycles in bla, 128 runs, 0 skips
[19:42:00] <BBB> 1177 dezicycles in bla, 256 runs, 0 skips
[19:42:00] <BBB> 1167 dezicycles in bla, 512 runs, 0 skips
[19:42:05] <Vitor1001> BBB: I'd say commit it :D
[19:42:40] <BBB> trying one more thing, then I'll commit I guess
[19:43:04] <kshishkov> BBB: in any case you can improve it later in SVN ;)
[19:43:43] <peloverde> Vitor1001: the Feb 2 e-mail? I must have missed it in the shuffle before FOSDEM, I'll give it another look
[19:44:07] <BBB> 1269 dezicycles in bla, 128 runs, 0 skips
[19:44:07] <BBB> 1260 dezicycles in bla, 256 runs, 0 skips
[19:44:07] <BBB> 1256 dezicycles in bla, 512 runs, 0 skips
[19:44:19] <BBB> so more dezicycles is bad right?
[19:44:24] <BBB> ok, kshishkov, yours is better :)
[19:44:30] <kshishkov> sorry
[19:44:36] <Vitor1001> peloverde: Yes, that one.
[19:45:14] <BBB> ohwell
[19:45:20] <BBB> at least I finished something
[19:45:29] <peloverde> But that only seems to handle the real part of the output
[19:45:52] <Vitor1001> No
[19:45:56] <Vitor1001> F_k is complex
[19:46:21] <Vitor1001> When I say "F_k + F_k-1" I mean a complex substraction
[19:46:42] <Vitor1001> The only thing that is not obvious is to find the complex part of F_0
[19:47:12] <Vitor1001> That you have to actually do sum_n X_n sin(...)
[19:47:46] <iive> BBB: the code you pasted. are you sure you haven't replaced == into = ?
[19:48:25] <iive> hum, my bad
[19:48:30] <kshishkov> no, it's all assignments
[19:49:04] <iive> i took it for (cache=n)?-n:0
[19:49:21] <iive> instead of cache= (n?-n:0)
[19:49:43] <iive> today is not my day.
[19:51:15] <BBB> :)
[19:51:19] <kshishkov> yes, go RE some codec
[19:51:26] <_av500_> Kovensky: i'm the somewhat tall guy in the back :)
[19:51:32] <BBB> have to do postfilter first
[19:51:36] <BBB> then a video codec or so
[19:52:04] <kshishkov> lol, there's _always_ speech codec postfilter left to be done
[19:53:31] <Dark_Shikari> BBB: WVP2!
[19:53:53] * kshishkov grins
[19:55:55] <Kovensky> _av500_: heh
[19:56:13] <_av500_> you might not have extrapolated that...
[19:57:09] <kshishkov> _av500_: okay, which of you disguised as a girl then?
[19:58:14] <_av500_> for i think she was genuine
[19:58:29] <_av500_> i think she was genuine
[19:58:33] <Kovensky> http://somafm.com/play/missioncontrol
[19:58:36] <kshishkov> genuine FFmpeg dev?!
[19:59:03] <_av500_> she does study cs, but i doubt she gave birth to a patch yet
[20:00:30] <kshishkov> my group at university had ~12 girls who got B.Sc. in CS
[20:01:01] <kshishkov> not that any of them actually _knew_ CS
[20:01:24] <Kovensky> lulz
[20:01:34] <peloverde> Vitor1001: I got it working
[20:02:02] <peloverde> but I'm still convinced there has fo be a clever way to game f[0].im
[20:02:06] <kshishkov> Kovensky: that's why I don't value our education
[20:02:23] * Kovensky doesn't value his education either
[20:02:33] <Vitor1001> peloverde: :D
[20:02:54] <Kovensky> I'm on supposedly the best college in town (scholarship), but I still haven't learned anything of value in it for the past 2 years other than calculus
[20:03:08] <Vitor1001> peloverde: I spent some time thinking about it, not really trivial :(
[20:03:17] <Dark_Shikari> Kovensky: that's not a very good college then
[20:03:29] <Dark_Shikari> I'm learning haskell \o/
[20:03:34] <Kovensky> I know
[20:03:34] <twnqx> >_>
[20:03:34] <Kovensky> lol
[20:03:41] <Kovensky> this semseter I'll be taught HTML <_<
[20:03:43] <Kovensky> and PHP >_>
[20:03:44] <Dark_Shikari> LOL
[20:03:45] <twnqx> time for an x264 port!
[20:03:48] <Dark_Shikari> what the fuck
[20:03:49] <Kovensky> then next semester CSS
[20:03:51] <Dark_Shikari> wait wait
[20:03:52] <twnqx> to haskell, i mean.
[20:03:58] <Dark_Shikari> that's not a CS program
[20:04:01] <Dark_Shikari> that's a web development program
[20:04:02] <Kovensky> no, it's not
[20:04:06] <Dark_Shikari> aka utterly useless bullshit
[20:04:16] <Kovensky> my course isn't exactly CS though, it's information systems
[20:04:17] <kshishkov> ours included Java EE
[20:04:21] <Dark_Shikari> which won't get you a job as anything other than a codemonkey
[20:04:27] <Dark_Shikari> earning minimum codemonkey wage
[20:04:30] <Kovensky> (one more reason I want that japan scholarship, I could at least study (or pretend to study) real CS)
[20:04:47] <Dark_Shikari> if a CS program has classes "on a language", it's not a CS program
[20:04:48] <kshishkov> Dark_Shikari: our companies prefer undergraduated codemonkeys, thank you
[20:04:53] <Kovensky> well
[20:04:58] <Kovensky> it's "Web Developent"
[20:05:08] <Kovensky> we had "Object Oriented Language" last semester, which was just Java
[20:05:29] <Dark_Shikari> even my "haskell" class is "programming languages"
[20:05:32] <Kovensky> several of the questions on the tests asked about supposed object oriented stuff that if I did answer thinking only of the basic OO concepts would be wrong
[20:05:32] <Dark_Shikari> where we learn to write parsers, compilers, etc
[20:05:38] <Dark_Shikari> and we just use haskell because it's bloody easy to write parsers in
[20:05:46] <Kovensky> or asked stuff that only existed in Java
[20:06:05] <kshishkov> Dark_Shikari: everything is bloody easy with Perl!
[20:06:08] <Kovensky> then before that it was "Structured Language", which taught C, but nobody learned shit because the professor sucked
[20:06:10] <Dark_Shikari> kshishkov: not as easy as haskell
[20:06:14] <Dark_Shikari> haskell has automatic pattern matching
[20:06:21] <Dark_Shikari> for any expression whatsoever
[20:06:25] <kshishkov> Kovensky: it was Pascal for us
[20:06:35] <_av500_> begin ftw
[20:06:38] <Kovensky> and even before "Structured Language" there was "Programming Techniques II", with Pascal
[20:06:41] <Dark_Shikari> EvaluateExpression Num ArithOp Num
[20:06:53] <Dark_Shikari> or EvaluateExpression Exp Arithop Exp
[20:06:56] <Kovensky> and before it "Programming Techniques", with some shitty language
[20:07:05] <Dark_Shikari> you can write an infix parser in about 3 lines of code
[20:07:06] <Kovensky> that was basically Pascal in portuguese
[20:07:18] <kshishkov> Kovensky: now, your program seems to be closer and closer to ours
[20:07:18] <Kovensky> considering how much I hate portuguese, you can guess how fun the classes where...
[20:07:25] <Kovensky> were*
[20:07:27] <Dark_Shikari> pascal :/
[20:07:40] <Kovensky> our C manuals were hillarious
[20:07:49] <Kovensky> our professor kept confusing C and Pascal syntax
[20:07:57] <kshishkov> well, we have an official Pascal translated into Russian, called "Ershov's [programming] language"
[20:08:15] <kshishkov> it's nice nobody remembers about it
[20:08:15] <Kovensky> I wonder if his actual code was like that, every single array of his would be short allocated and the 0th element would always be unused
[20:08:50] <Dark_Shikari> pascal/C strings are the best
[20:08:51] <Kovensky> some of my colleagues actually complained when I told them what was the Right Thing, saying that if it's in the manual it's because it's like that ._.
[20:08:55] <Dark_Shikari> i.e. lenght header + zero-terminated
[20:09:39] <_av500_> Dark_Shikari: also 64byte cache line padded?
[20:09:43] <Kovensky> my C professor also frequently wrote code that either didn't compile at all or had blatant errors
[20:09:52] <kshishkov> DOS strings - dollar sign terminated!
[20:09:53] <Dark_Shikari> my PLs professor confuses haskell and ML syntax
[20:09:55] <Dark_Shikari> but he admits it
[20:09:58] <Dark_Shikari> and they're practically the same anyways
[20:10:03] <Kovensky> there was one time where he opened Dev-C++ and wrote an entire program in Pascal
[20:10:10] <Kovensky> (yes we are forced to use Dev-C++)
[20:10:45] <kshishkov> it was Borland C++ 4.5 for us
[20:10:54] <Kovensky> heh
[20:11:05] <Kovensky> I wonder if I can convince the new manager to at least pretend to do stuff right <_<
[20:11:27] <Kovensky> also see if they finally can give us direct MSDNAA access instead of asking the manager for every single key that we want
[20:11:55] * _av500_ has to remember to renew his msdnaa....
[20:12:10] <Kovensky> the manager doesn't even have the keys that I want :/
[20:12:13] <Kovensky> namely win7 ultimate
[20:12:22] <_av500_> yep, for that ;)
[20:12:32] <Kovensky> the college's MSDNAA subscription does give win7 access though
[20:12:40] <Kovensky> he had win7 home basic keys
[20:12:48] <superdump> Compn: is the archive dir the dir with all the symlinks?
[20:13:07] <superdump> i.e. http://samples.mplayerhq.hu/archive/
[20:15:36] * kshishkov has managed to live without MSDN even when he programmed ActiveX controls using VfW
[20:16:21] <kshishkov> *yawn*
[20:41:44] <Vitor1001> Kovensky: Just for curiosity, where is your college?
[20:42:55] <Kovensky> northeastern brazil
[20:43:15] <Kovensky> they have much better colleges in the southeast, which is pretty much the only part of the country that pretends to be developed
[20:43:24] <Vitor1001> Are u brazilian?
[20:43:24] <ramiro> Kovensky: Vitor1001 is brazilian too =)
[20:43:54] <Kovensky> o rly
[20:44:13] <Kovensky> and yes, I'm stuck in sao luis =p
[20:44:21] * Kovensky was born in salvador though
[20:44:24] <ramiro> do maranhao
[20:44:31] <ramiro> Kovensky: doesn't make it much better, hehehe
[20:44:50] * Kovensky wonders why must everyone append "do maranhao" to "sao luis"
[20:44:57] <Vitor1001> Nice...
[20:44:58] <Kovensky> not like there is any other "sao luis" that you should know about
[20:45:16] <Vitor1001> I was there not so long ago, to go to the lençois
[20:45:42] <Vitor1001> Really nice, but one should run away of the CVC (a tourist company) trap
[20:46:19] * Vitor1001 is from SP
[20:46:36] * Kovensky has lived in several parts of the country
[20:46:42] <ramiro> Vitor1001: my sister used to work for cvc
[20:46:45] * Kovensky got stuck in sao luis since 2000
[20:46:47] <ramiro> but she says the same thing...
[20:47:55] <Vitor1001> ramiro: Nothing against it, but for lençois they have the "10 days at lençois" that is actually 3 days at lençois and 7 at sao luis
[20:49:33] <Vitor1001> Kovensky: I wonder why there is not more people from the NE that do the univ. entrance exams in the southeast...
[20:50:07] <ramiro> well, ufpe in recife doesn't seem to be that bad, and it's closer.
[20:50:59] <Vitor1001> it's fun to go a few thousand km away from home sometimes ;)
[20:51:08] <Kovensky> I know
[20:51:58] <Kovensky> it's mostly that I'm poor (or I wouldn't be stuck here)._.
[20:52:02] <Kovensky> +
[20:53:02] <Vitor1001> I see... I know that at least in SP it is pretty hard to get university housing...
[20:53:25] <Vitor1001> And I don't say "decent university housing" because there is no such a thing
[20:54:01] <ramiro> Vitor1001: from what I've been told, only "uncontroversial" people (that don't party and don't protest) get university housing in usp. basically only crentes.
[20:54:27] <Vitor1001> ramiro: No, you just have to be poor
[20:54:31] <Vitor1001> I mean _really_ poor
[20:54:56] <Vitor1001> It's paperwork
[20:55:13] <Vitor1001> there will be no "do you like to party?" question in the forms
[20:55:25] * Kovensky is not poor enough for most stuff
[20:55:48] <Kovensky> specially since the main measure of poverty here is the power bill, and mine is pretty high
[20:55:57] <Kovensky> R$400, R$500
[20:56:44] <ramiro> Kovensky: incredible, you manage to spend even more than me.
[20:57:12] <Vitor1001> Kovensky: I spend less than that, and I need some serious heating here in paris!
[20:57:22] <ramiro> and I have a quad core running FATE boxes on the living room and a broken fridge that eats electricity.
[20:58:51] <CIA-17> ffmpeg: siretart * r21758 /branches/ (0.5 0.5/libavcodec/snow.c):
[20:58:51] <CIA-17> ffmpeg: Make sure the block array is of the correct size.
[20:58:51] <CIA-17> ffmpeg: This might have been exploitable.
[20:58:51] <CIA-17> ffmpeg: backported r18393 by michael
[20:58:52] <Kovensky> I have 5 Sempron 2800+ boxes + my craptop + tv + fridge + microwave
[20:59:02] <Kovensky> we run an improvised lan house @ home ._.
[20:59:02] <Vitor1001> no AC?
[20:59:13] <Kovensky> no AC, and to install one we'd have to rebuild the house
[20:59:17] <thresh> heating in paris? stop joking :-)
[20:59:46] <Vitor1001> thresh: lol, where are you?
[20:59:57] <Kovensky> also, for heating, just run f at h, compile something, transcode, etc
[21:00:00] <Kovensky> :)
[21:00:06] <thresh> Moscow Area, sadly
[21:00:50] <Vitor1001> thresh: Yeah, I can imagine some serious cold there
[21:01:39] * Kovensky wants to escape to yurop or jewpan
[21:01:40] <BBB> Vitor1001, I'm writing specs for a little, then I'll commit
[21:01:53] <Vitor1001> BBB: nice!
[21:01:56] <Vitor1001> congrats!
[21:02:05] <thresh> the average monthly pay for public heating services and other stuff like water rarely exceeds 100 EUR here
[21:02:07] <Vitor1001> Pretty complex for a first decoder :)
[21:02:45] <Vitor1001> thresh: That quite a bit, supposing gas is cheap there
[21:03:03] <Vitor1001> I pay around 800 EUR / year
[21:03:15] <thresh> well, that's for a 3-room flat with 4 people living in it
[21:03:33] <Vitor1001> thresh: But then, it is a Paris sized appartment for 2
[21:03:35] <thresh> for 1-room one, i'd say around 40 EUR
[21:04:13] <thresh> however it is pretty warm here when it's -30°C outside
[21:04:19] <thresh> provided you have proper windows of course
[21:04:24] <CIA-17> ffmpeg: siretart * r21759 /branches/ (0.5 0.5/libavcodec/mlpdec.c):
[21:04:24] <CIA-17> ffmpeg: Fix crash in MLP decoder due to integer overflow.
[21:04:24] <CIA-17> ffmpeg: Probably only DoS, init_get_bits sets buffer to NULL, thus causing a
[21:04:24] <CIA-17> ffmpeg: NULL-dereference directly after.
[21:04:24] <CIA-17> ffmpeg: backport r21426 by reimar
[21:04:46] <_av500_> thresh: windows7?
[21:05:17] <Kovensky> Vitor1001: how did you escape the banana republic
[21:05:40] <Vitor1001> A partnership between USP and a french university
[21:05:49] * Kovensky should try and escape to sweden
[21:06:03] <Kovensky> nice climate and cool music
[21:06:08] <Kovensky> plus good economy
[21:06:10] <thresh> nice climate??
[21:06:15] <thresh> it's cold out there
[21:06:21] <Honoome> that's good
[21:06:29] <Vitor1001> Kovensky: If you are into studing, you can try doing a master of science and them a PhD abroad
[21:06:30] <Kovensky> thresh: compare to the 45C in Santos, or 33C in here
[21:06:32] <Honoome> you can overclock more
[21:06:53] <thresh> Kovensky: well, yeah, oz is a bit warm
[21:07:04] <thresh> although i'd prefer melting rather than icing
[21:07:24] <mru> wizards like it warm?
[21:07:35] <mru> or was that lizards?
[21:08:09] <mru> the oz govt seems to be controlled by the lizard army at least
[21:08:16] <Kovensky> Vitor1001: my current plan is trying for a scholarship in japan
[21:08:44] <Honoome> mru: take us to your lizard?
[21:09:07] <Kovensky> oz?
[21:09:38] <Vitor1001> Kovensky: could work
[21:10:17] <Kovensky> Vitor1001: what will you do once the partnership finishes btw?
[21:10:44] <Vitor1001> Kovensky: It has finished a long time ago... That was for college, I'm finishing my PhD now.
[21:10:51] <Kovensky> oic
[21:11:19] * Kovensky dislikes his mother language and is currently studying japanese
[21:11:34] <Kovensky> studying in japan would be ideal
[21:11:44] <mru> s/ideal/essential/
[21:12:05] <Kovensky> mru: well, not for learning japanese; I could just go to Liberdade
[21:12:25] <Kovensky> (an almost completely japanese neighborhood in Sao Paulo)
[21:12:27] <Vitor1001> Kovensky: It is still quite far from sao luis
[21:12:32] <Kovensky> Vitor1001: true =p
[21:14:46] <Kovensky> it's been raining the whole day in here and it's still freaking hot q_q
[21:15:57] * Kovensky foobar2000 (v1.0): Amon Amarth [2008 Twilight of the Thunder God #1.04/1.10] Where Is Your God? [0:22/3:11] 281kbps MP3 VBR <-- example of cool swede music
[21:22:30] <peloverde> Amon Amarth is pretty solid
[21:26:08] <superdump> is it metal?
[21:26:14] <superdump> it sounds like a metallish name
[21:26:20] <Kovensky> melodic death metal about vikings
[21:26:39] <Kovensky> Amon Amarth is an alternative name for Mordor in LoTR
[21:26:42] <Kovensky> LotR*
[21:26:45] * Kovensky is case sensitive
[21:27:34] <Kovensky> (well, or so wikipedia says)
[21:39:28] <_av500_> ramiro: forget the pools and use heap
[21:39:40] <_av500_> that saves you from having to track each buffer size you use
[21:40:53] <_av500_> note that you have to change one Memory_alloc() param for heap
[21:41:20] <superdump> Kovensky: aha, i thought i recognised it
[21:42:01] <_av500_> ramiro: also, the loadmodules.sh values are way oversized, they can be cut down a lot
[21:44:56] <ramiro> _av500_: doesn't the h264 decoder allocate that memory internally? do I have control on how it allocates it?
[21:47:58] <ramiro> _av500_: and I don't know how much I can cut them down. as I said on the list the documentation says the sizes, but not how many buffers of each are needed.
[21:54:57] <CIA-17> ffmpeg: cehoyos * r21760 /trunk/ffmpeg.c:
[21:54:57] <CIA-17> ffmpeg: Remove recording_time check which is no longer necessary after r21687.
[21:54:57] <CIA-17> ffmpeg: Patch by Wolfram Gloger, wmglo A dent D med D uni-muenchen D de
[21:59:50] <_av500_> ramiro: what cpu are u on?
[22:00:05] <ramiro> _av500_: dm365
[22:00:42] <_av500_> hm that might differ from arm plus dsp setup...
[22:11:02] <_av500_> ramiro: do you compile code that allocs cmem?
[22:11:09] <_av500_> or is that object code you use?
[22:12:24] <mru> there's a difference?
[22:12:38] <_av500_> well, source is easier to change :)
[22:12:52] <mru> then you need a better hex editor
[22:12:59] <_av500_> i dont know how dm36x sdks do it
[22:13:05] <ramiro> _av500_: how will I get contiguous memory from the heap without cmem pools though?
[22:13:21] <_av500_> you can setup cmem to use a heap instead of pools
[22:13:38] <_av500_> in that case it has one big heap and allocsfrom there
[22:13:45] <_av500_> much like malloc
[22:13:50] <mru> you can have a mix too
[22:13:53] <_av500_> yes
[22:14:04] <_av500_> pro is that you dont need to worry aboutbuffer sizes
[22:14:14] <_av500_> con is that ti is afraid the pool will fragment...
[22:14:15] <ramiro> and con?
[22:14:16] <mru> if you have only a few different sizes pools are better
[22:14:21] <mru> faster and no fragmentation
[22:14:31] <mru> pools don't fragment, heaps do
[22:14:35] <_av500_> but we use a heap since 4 ys and never saw any
[22:14:55] <mru> if your usage pattern clears it out from time to time you're fine
[22:15:11] <_av500_> we could not define a set of pools that matches 10 different codecs
[22:15:25] <ramiro> another problem I'm having is the oom killer killing vlc
[22:15:35] <_av500_> ;)
[22:15:36] <mru> _av500_: do you free everything after playing a file?
[22:15:54] <ramiro> the idle system with top running syays some 68Mb of memory is being used. with vlc uses up to 20% memory it's killed
[22:15:57] <_av500_> mru: most of it
[22:16:10] <ramiro> how do I find out who's hogging all those 68Mb on the idle system?
[22:16:21] <mru> ps?
[22:16:23] <_av500_> mru: yes, heap state after play is more or less the same as before
[22:16:32] <_av500_> its not like you star a video while one is playing...
[22:16:35] <_av500_> start
[22:16:48] <mru> what, no pip?
[22:16:49] <_av500_> that would fragment, but has little real world use...
[22:17:06] <ramiro> if I add up the memory from all the programs from ps or top it goes to about 10%... much different from the 90%
[22:26:46] <mru> Dark_Shikari: ping
[22:26:53] <Dark_Shikari> pong
[22:27:12] <mru> do you still have that collection of h264 test streams up somewhere?
[22:27:36] <Dark_Shikari> http://www.mediafire.com/download.php?mi2zjoynymi
[22:29:45] <mru> thanks
[22:33:36] <CIA-17> ffmpeg: michael * r21761 /trunk/libavformat/iv8.c: Fix timestamps.
[22:34:30] <mru> is that supposed to fix fate?
[22:38:51] <BBB> bleh
[22:38:54] <BBB> enough spec writing for today
[22:40:29] <Honoome> BBB: what about rspec writing? :D I could use someone to do that on a few packages :P
[22:42:57] <BBB> not my thing
[22:43:00] <BBB> ;)
[22:43:32] <Honoome> not masochist enough, uh?
[22:52:59] * /join #ffmpeg-devel ...
[22:53:01] *** TOPIC: Welcome to the FFmpeg development channel. That is, development of FFmpeg, not using FFmpeg, nor libav*. | Users should redirect their questions to #ffmpeg | FFmpeg 0.5 has been released! | this chat is now publicly logged.
[22:53:01] *** TOPICINFO: Compn, 1264791848
[22:53:06] <BBB> no more comments on the wmavoice codec?
[22:53:07] <BBB> going once...
[22:53:09] <BBB> going twice...
[22:53:20] <Compn> you ran the fuzzer on it ?
[22:53:27] <BBB> what is the fuzzer?
[22:53:32] <mru> and valgrind?
[22:53:38] <Compn> fuzzed input to test if it crashes or leaks memory
[22:53:38] <DonDiego> tools/trasher ?
[22:53:45] <BBB> oh you guys suck, now you come up witht hat?
[22:53:45] <Compn> zzuf
[22:53:47] <Compn> lol
[22:53:59] <BBB> I don't do valgrind, no linux here
[22:54:00] <Compn> BBB : isnt it in developer documentation ?
[22:54:13] <Compn> if its not, DonDiego should put it in there
[22:55:03] <ramiro> BBB: no linux? what do you develop on?
[22:55:30] <Compn> we can guess he devels on windows :P
[22:55:45] <DonDiego> no, os x
[22:55:46] <mru> if no linux, get linux
[22:56:00] <ramiro> IIRC valgrind now claims to work on osx
[22:59:32] <BBB> ramiro: macosx
[23:00:00] <BBB> let me check
[23:00:18] <BBB> $ sudo port install valgrind
[23:00:18] <BBB> Password:
[23:00:18] <BBB> Error: Port valgrind not found
[23:00:19] <BBB> no
[23:00:47] <Compn> sudo port update? :P
[23:01:04] <ramiro> 19 August 2009: valgrind-3.5.0, for X86/Linux, AMD64/Linux, PPC32/Linux, PPC64/Linux and X86/Darwin (Mac OS X) is available.
[23:01:06] <Compn> Valgrind's Mac OS X support is now part of Valgrind's main development trunk.
[23:01:31] <DonDiego> http://trac.macports.org/browser/trunk/dports/devel/valgrind/Portfile
[23:02:16] <BBB> why isn't it here then?
[23:02:20] <BBB> maybe time to update my ports
[23:03:34] <BBB> I did just download several more files from random locations on internet
[23:03:36] <BBB> and they all play
[23:03:46] <BBB> I even found a second sample file using WMAPro-in-WMAVoice
[23:03:52] <BBB> maybe I should implement that some day
[23:04:34] * Compn runs afk after botching up BBB's attempted commit
[23:04:48] <BBB> damn you :)
[23:05:21] <BBB> installing valgrind will take ages
[23:05:56] <DonDiego> try valgrind after committing then..
[23:06:17] <BBB> no, I'll valgrind it tonight
[23:06:19] <BBB> gives me something fun to do
[23:06:28] <BBB> did zhentan feng come online today?
[23:07:20] <BBB> checking for a supported OS... ok (darwin10.2.0)
[23:07:20] <BBB> checking for the kernel version... unsupported (10.2.0)
[23:07:20] <BBB> configure: error: Valgrind works on Darwin 9.x (Mac OS X 10.5)
[23:07:27] <BBB> valgrind is majorly confused, I'm running 10.6
[23:07:59] <BBB> this won't work
[23:08:11] <BBB> I'll do fuzz thingy tonight instead
[23:08:15] <BBB> how does trasher work?
[23:08:25] <BBB> makes random modifications in files
[23:08:25] <BBB> ?
[23:13:59] <lu_zero> BBB: you need valgrind svn I'm afraid
[23:14:18] <lu_zero> your os is newer than the one supported ^^;
[23:14:34] <BBB> I'll valgrind after commit, as diego suggested
[23:14:42] <BBB> I'm gonna fuzz tonight and see what that does
[23:14:44] <BBB> should be fun
[23:14:45] <peloverde> BBB: I don't know about trasher but here is a zzuf tutorial: http://caca.zoy.org/files/zzuf/zzuf-20070225.pdf
[23:16:22] <astrange> you need the patch from https://bugs.kde.org/show_bug.cgi?id=205241
[23:19:58] <BBB> damn these developers that can't write portable software
[23:20:04] <BBB> even zzuf doesn't compile straight from ports
[23:20:44] <astrange> valgrind can't be written portably
[23:21:07] <peloverde> http://caca.zoy.org/attachment/wiki/zzuf/zzuf-osx-0.13.tar.gz
[23:22:15] <mru> most of it can
[23:22:34] <BBB> \o/
[23:22:38] <BBB> thanks peloverde
[23:22:40] <BBB> will play tonight
[23:22:47] * BBB -> home
[23:41:28] <CIA-17> ffmpeg: mru * r21762 /trunk/configure: configure: make mdct and rdft select fft and update other deps
[23:41:28] <CIA-17> ffmpeg: mru * r21763 /trunk/configure: configure: add missing mdct deps
[23:41:29] <CIA-17> ffmpeg: mru * r21764 /trunk/libavcodec/fft.c: Fix build with --disable-mdct
[23:41:30] <CIA-17> ffmpeg: mru * r21765 /trunk/configure:
[23:41:30] <CIA-17> ffmpeg: ffplay depends on rdft
[23:41:30] <CIA-17> ffmpeg: Spotted by Ramiro.
[23:58:40] <CIA-17> ffmpeg: mru * r21766 /trunk/configure: configure: require --arch and --target-os when cross-compiling

More information about the FFmpeg-devel-irc mailing list