[FFmpeg-devel-irc] IRC log for 2010-08-20

[irc at mansr.com]

[05:04:52] <thresh> moroning
[05:16:56] <saintdev> is there a FHT anywhere in ffmpeg?
[05:24:35] <saintdev> ahh. the FHT is just an all-real analog of the FFT
[05:25:37] <saintdev> I wonder if I can just use the MDCT coefficients for this, like vorbis does?
[05:26:26] <saintdev> question the second, do we have access to an all-real FFT?
[05:28:29] <kshishkov> we have RDFT
[05:28:46] <kshishkov> it's used for Bink audio, FFplay and somewhere else
[05:28:57] <saintdev> Real-DFT, I take it?
[05:32:07] <saintdev> I wonder if that would be close enough. Guess I should try using MDCT coeffs first.
[07:28:56] <astrange> http://gmailblog.blogspot.com/2010/08/use-linux-now-you-can-video-chat-too.html
[07:33:45] <kshishkov> okay, what codecs are there?
[07:35:53] <astrange> i was hoping someone else would say
[07:36:16] <av500> bink?
[07:36:40] * kshishkov sends plushy Jar-Jar Binks to av500
[07:36:54] * av500 uses them for target practice
[07:37:08] <kshishkov> astrange: strings on main decoder say something
[07:37:14] <kshishkov> hmm, Lmi H.264 encoder
[07:37:40] <av500> what is a .deb anyway?
[07:38:10] <kshishkov> g.711, PCM, isac/isaclc, iLBC, G.722, GSm, Speex,
[07:38:28] <kshishkov> av500: installation package for Debian/Ubuntu. Tar xzf should unpack it
[07:38:57] <kshishkov> and H.263
[07:39:35] <kshishkov> H.264 SVC? hmmm
[07:40:45] <thresh> tar wouldnt unpack it
[07:40:52] * kshishkov makes av500 wear T-shirt "I liked Wesley character from TNG!" while he's not watching
[07:40:59] <thresh> or would it?
[07:41:06] <av500> ar -x
[07:41:12] <thresh> i always that it is ar+cpio
[07:41:13] <thresh> yeah
[07:41:18] <thresh> +thought
[07:41:28] <kshishkov> whatever, it's standard archive anyway
[07:41:33] <kshishkov> (unlike RPM)
[07:41:48] <thresh> ORLY
[07:42:04] <av500> great, rpm vs deb war!!!
[07:42:15] <kshishkov> thresh: да, орлы
[07:42:38] * kshishkov is long time Fedora user
[07:43:12] <kshishkov> personally I don't care about package format much, be it .deb, .rpm or .tgz
[07:43:33] <thresh> rpm2cpio ftw
[07:43:50] <kshishkov> I know
[07:44:12] * kshishkov preferred Midnight Commander to browse them all anyway
[07:44:23] * av500 found "Hello, world!" in libnpgtpo3dautoplugin.so, prepared to sue google
[07:46:24] <kshishkov> av500: use Word 2000 to write letter, Clippy will be happy to assist you
[07:47:02] <av500> hmm, word2000 is 1994 versions after word6?
[07:47:15] <av500> it must have improved a lot then...
[07:47:17] <kshishkov> nope
[07:48:14] <kshishkov> they got word6, then word 95-97, then word 2000, then 16-bit overflow so they settled with XP
[08:15:52] <superdump> kshishkov: i tried some trocadero from an OKQ8 yesterday
[08:16:03] <superdump> it was a bit too fizzy straight out of the bottle
[08:16:09] <superdump> but the flavour was quite refreshing
[08:24:24] <kshishkov> superdump: you haven't tried Julmust/PÃ¥skmust then
[08:24:44] <superdump> must?
[08:24:48] <superdump> what's that?
[08:24:58] <mru> a drink
[08:24:59] <mru> or two
[08:25:03] <kshishkov> http://en.wikipedia.org/wiki/Julmust
[08:26:38] <superdump> yeah
[08:26:45] <av500> malzbier?
[08:26:50] <superdump> lol @ coca cola AB
[08:26:59] <superdump> just to produce coca cola julmust
[08:27:06] <superdump> because it impacts on their sales so much
[08:27:23] * superdump makes a note to check the brand of the julmust before buying
[08:27:38] <superdump> i'm not a big fan of carbonated soft drinks though
[08:27:42] <kshishkov> that article says it went out of production anyway
[08:27:42] <superdump> i prefer still stuff
[08:27:55] * kshishkov hates still water after visiting Denmark
[08:28:35] <av500> kshishkov: dont drink the sea water!
[08:29:40] <kshishkov> av500: I don't.
[09:45:31] <av500> my pogoplug compatible runs ffmpeg :)
[09:48:54] <mru> my sheeva is back in action
[09:48:59] <mru> free of its shell
[09:51:49] * mru finds a horrible design bug in swscale
[09:51:52] <mru> s/a/another/
[09:53:24] <kshishkov> is that bug called "swscale"?
[09:54:18] <kshishkov> well, it's originated from borrowed libmpeg2dec code in MPlayer so it's all hacks
[09:54:42] <kshishkov> too bad that an effort to rewrite it failed
[10:12:50] <mru> seriously though, this needs to be fixed
[10:15:58] <av500> mru: http://ffmpeg.pastebin.com/yQj7E5EG
[10:16:30] <mru> yes?
[12:19:32] <lu_zero> mru: rewritten how?
[12:21:19] <av500> in c++...
[12:22:44] <lu_zero> av500: in haskell you mean
[12:24:00] <av500> pascel?
[12:25:45] <lu_zero> pascal?
[12:25:51] <lu_zero> erlang!
[12:25:59] <lu_zero> let's rewrite all in erlang!
[12:34:09] <kshishkov> lu_zero: may I remind you where Erlang was developed and it makes sense for FFmpeg?
[12:34:46] <KotH> let's do it in scheme
[12:35:10] <kshishkov> ELisp is more popular
[12:37:28] <thresh> this what google gave me for 'people who code on erlang': http://1.bp.blogspot.com/_bXDgUoT7V7o/Sk4AeAT0K7I/AAAAAAAAANA/5lDV0iMS_V8/s400/newkids3.jpg
[12:37:49] <thresh> i could never be as awesome as erlang programmers :(
[12:37:51] <av500> new coders on the block?
[12:37:58] <kshishkov> yep, looks better than people who code in PHP
[12:38:08] <kshishkov> err, s/people/monkeys/
[12:39:51] <Tjoppen> I've wondered why swscale isn't built to be able to cobble together transforms on its own. it seems rather "mechanical" atm
[12:41:03] <KotH> thresh: i remember when people were walkin dressed like that on the streets :)
[12:41:05] <cartman> thresh: lol
[12:41:27] <thresh> KotH: I do too. Last year.
[12:41:32] <KotH> lol
[12:41:44] <KotH> did you fall in a space-time virtex?
[12:42:06] <thresh> yes, isnt it the translation of Russia in european languages?
[12:43:14] <thresh> seriously, jackets like red one are hugely popular here
[12:43:27] <thresh> among other 'I AM A RACER' like ones
[12:46:26] <kshishkov> малиновые пиджаки, золотые цепи
[12:47:40] <thresh> kshishkov: это позапрошлый год :)
[12:48:11] <kshishkov> thresh: а у конкретных пацанов оно всегда в моде, реальна!
[13:00:54] <mru> lu_zero: I said fixed, not (necessarily) rewritten
[13:01:16] <kshishkov> sometimes it's the only appropriate fix
[13:01:19] <mru> right now there's code doing one unaligned load per pixel
[13:01:24] <mru> obviously a bad idea
[13:04:28] <mru> KotH: space-time virtex? some new all-powerful fpga?
[13:04:29] <kshishkov> hmm, is that why that one test fails on many systems making FATE more yellow than it should be?
[13:04:34] <mru> yes
[13:05:14] <kshishkov> mru: probably in his case it was "vortex" typo. Or could be an abbreviation for "virtual technology"
[13:05:15] <KotH> mru: juup.. all powered by an improbability field
[13:14:54] <lu_zero> KotH: I want one!
[13:24:39] <KotH> lu_zero: they're available, quite cheaply, at CERN :)
[13:25:01] <av500> so, ask mru sister for some
[13:25:36] <kshishkov> av500: I thought you're going to ask her for pet llama offspring for your kids
[13:25:52] <mru> different sister
[13:26:36] <lu_zero> mru: you don't have other siblings?
[13:26:38] * KotH doesnt think that non-plush lamas are a good gift for kids
[13:26:43] <av500> kshishkov: I would love to see pet llama offspring going through the CERN black hole
[13:27:01] <av500> I guess it will take the whole family to pull that off
[13:28:10] <lu_zero> av500: just to have a quantum roast llama?
[13:28:46] <av500> lu_zero: yes, goes well with the pumpkin soup
[13:29:14] <lu_zero> uhmm
[13:29:20] <lu_zero> pumpkin soup
[13:29:20] <KotH> is today no-ffmail day or something?
[13:29:21] <lu_zero> yum
[13:29:30] * lu_zero is 1/2 ill
[13:29:44] * KotH got worried when he had a look at todays mail stats
[13:29:51] <lu_zero> and my crazy partners want to go camping in the mountains
[13:30:01] <lu_zero> =E
[13:30:14] <KotH> lu_zero: do i know them? :)
[13:30:49] <lu_zero> I doubt
[13:30:58] <KotH> do they know me?
[13:31:00] <av500> crazy mountain folks
[13:31:05] <av500> they should
[13:31:15] <av500> you are from montain country too
[13:31:36] <lu_zero> one is Puria and you might know him
[13:31:38] * KotH will be in the mountains this evening
[13:31:43] <kshishkov> av500: don't say that to somebody with nickname abbreviated from "Kurd of the Helvetia"
[13:31:49] <KotH> to translate japanese <-> swiss german
[13:31:58] * thresh will be in mountains next year, if lucky
[13:32:05] <KotH> thresh: you'll be in .ch?
[13:32:09] <lu_zero> the other is Alessandro Molina
[13:32:21] <thresh> KotH: no idea, probably not (too expensive)
[13:32:37] * kshishkov was in Basle around a month ago
[13:32:42] <lu_zero> 15:31 < KotH> to translate japanese <-> swiss german
[13:32:45] <lu_zero> eh?
[13:32:53] <KotH> lu_zero: the name allesandro molina rings a bell, but i cannot tell from where i might know him
[13:32:54] <thresh> Bulgaria, France, Switzerland are sure guesses for next year... have to decide on two other vacation dates still
[13:33:14] <KotH> lu_zero: you know, people hire other people to translate between different languages
[13:33:29] * KotH got once hired to translate turkish <-> japanese at a wedding in zürich
[13:33:37] <lu_zero> is the mix sounding strange
[13:33:51] <KotH> strange people attract strange mixtures
[13:33:52] <av500> KotH: also during wedding night?
[13:34:03] <av500> sounds fun
[13:34:07] <KotH> av500: there was no night left after the party :-)
[13:34:21] <lu_zero> eheheh
[13:34:33] <KotH> i think, it was 5am, when i finaly left
[13:34:47] <lu_zero> KotH: japanese people drinking turkish alcool?
[13:35:25] <kshishkov> lu_zero: yes, the best turkish alcohol blessed by imam
[13:35:55] <lu_zero> kshishkov: alcohol is part of the turkish tradition...
[13:36:04] <KotH> lu_zero: juup
[13:36:10] <kshishkov> lu_zero: go tell that to Russians
[13:36:38] <av500> that anis flavored stuff is more like an anti alcohol treatment to me
[13:37:20] <thresh> kshishkov: alcohol is not a tradition here
[13:37:25] <thresh> it's more of a everyday drink
[13:37:28] <lu_zero> kshishkov: those are nearly edible fuel
[13:38:19] <kshishkov> thresh: well, what about drunken fight on wedding?
[13:38:52] <lu_zero> O_o?
[13:41:44] <KotH> av500: did you do the mistake and drink raki pure?
[13:41:59] <av500> instead of?
[13:42:07] <av500> raki on the rocks?
[13:42:15] <thresh> kshishkov: I will attend wedding tomorrow, so will check on this one!
[13:42:18] <kshishkov> av500: ask iive about Gabrovo style of raki
[13:42:35] <thresh> rakia owns
[13:42:52] * thresh still has two bottles in a bar
[13:43:19] * kshishkov points out that Ukraine owns more - there is no word "vodka" in Ukrainian
[13:43:47] <thresh> so Ukrainians not only steal gas from us, you steal words too!
[13:45:18] <kshishkov> thresh: no, it's other languages stole your word, Ukrainian simply doesn't have it (maybe it was stolen by Russians?)
[13:48:05] <lu_zero> kshishkov: what's the traditional drink then?
[13:48:23] <thresh> russian blood
[13:49:14] <kshishkov> horilka
[13:50:35] <kshishkov> lu_zero: and Russian traditional drink is "anything that burns"
[13:50:46] <thresh> true
[13:55:13] <lu_zero> ah...
[14:08:00] <KotH> av500: raki is diluded 1:3 and drunk with a cup of yoghurt to cool the burn :)
[14:08:19] <KotH> av500: only the die hard alcoholics drink it pure
[14:11:50] <av500> ah, so thats why you buy so much yoghurt...
[14:15:50] <KotH> nah.. that's for ayran :)
[14:20:44] <CIA-93> ffmpeg: stefano * r24842 /trunk/libavfilter/avfilter.h:
[14:20:44] <CIA-93> ffmpeg: Cosmetics: add an empty newline between the function description and
[14:20:44] <CIA-93> ffmpeg: the list of @params.
[14:20:44] <CIA-93> ffmpeg: Improve consistency and possibly enhance readability.
[14:22:55] <lu_zero> ayran is wonderful
[14:23:03] <lu_zero> (with some mint leaves)
[14:24:29] <cartman> lu_zero: thats yummy indeed
[14:26:38] <KotH> salut Rathann|BlackHole
[14:27:48] <Rathann|work> salut KotH
[14:28:52] <kierank> black holes have internet?
[14:29:04] <av500> at cern, yes
[14:29:08] <av500> they invented it
[14:29:26] <Rathann|work> and the increased gravity makes it go faster, too
[14:29:31] <av500> al gore came out of one and brought it with him afaik
[14:42:10] <CIA-93> ffmpeg: stefano * r24843 /trunk/libavfilter/avfilter.c:
[14:42:10] <CIA-93> ffmpeg: Only print the pointer to the first plane in ff_dprintf_picref().
[14:42:10] <CIA-93> ffmpeg: To display the other planes is usually not useful and add noise to the
[14:42:10] <CIA-93> ffmpeg: output.
[14:42:10] <CIA-93> ffmpeg: stefano * r24844 /trunk/libavfilter/avfilter.c: Make ff_dprintf_picref() print video properties only if available.
[14:42:12] <CIA-93> ffmpeg: stefano * r24845 /trunk/libavfilter/avfilter.c:
[14:42:12] <CIA-93> ffmpeg: Extend ff_dprintf_picref() to make it print video interlaced and
[14:42:12] <CIA-93> ffmpeg: top_field_first information.
[15:11:54] <BBB> punpcklbw 1(%1), %%mm0 \n\t <- what does that 1(%1) mean?
[15:12:28] <mru> %1 is the second operand to the asm block
[15:12:37] <Dark_Shikari> 1(%1) means [%1+1]
[15:12:43] <BBB> ah
[15:12:47] <BBB> inline asm is weird
[15:12:52] <Dark_Shikari> duhr
[15:12:53] <mru> different syntax
[15:12:56] <peloverde> x86 is weird
[15:13:03] <mru> neither is more weird than the other really
[15:13:08] <Dark_Shikari> I don't think x86 is at fault for gcc inlining sucking
[15:16:51] <CIA-93> ffmpeg: stefano * r24846 /trunk/libavfilter/ (avfilter.c internal.h):
[15:16:51] <CIA-93> ffmpeg: Rename ff_dprintf_picref() to ff_dprintf_ref().
[15:16:51] <CIA-93> ffmpeg: The function is going to be used to represent also audio data.
[15:16:51] <CIA-93> ffmpeg: stefano * r24847 /trunk/libavfilter/avfilter.c:
[15:16:51] <CIA-93> ffmpeg: Make ff_dprintf_ref() print the information related to the referenced
[15:16:51] <CIA-93> ffmpeg: AVFilterBuffer.
[15:16:52] <CIA-93> ffmpeg: stefano * r24848 /trunk/libavfilter/avfilter.c: Cosmetics: merge two lines in ff_dprintf_ref().
[15:16:52] <CIA-93> ffmpeg: stefano * r24849 /trunk/libavfilter/avfilter.c: Make ff_dprintf_ref() print audio related information if available.
[15:19:41] <roozhou> is it possible to put label to put label in offset? e.g. label1(%1)
[15:49:04] <CIA-93> libswscale: ramiro * r31984 /trunk/libswscale/utils.c:
[15:49:04] <CIA-93> libswscale: fix anonymous memory mapping for NetBSD
[15:49:04] <CIA-93> libswscale: mmap() with MAP_ANONYMOUS requires the file descriptor to be -1 in NetBSD.
[15:49:04] <CIA-93> libswscale: Linux just ignores this parameter.
[15:49:04] <CIA-93> libswscale: Patch by Grant Carver <grantc at cat dot co dot za>
[16:35:31] <CIA-93> ffmpeg: stefano * r24850 /trunk/libavcore/imgutils.h: Add missing period in av_fill_image_max_pixstep() doxy.
[16:47:53] * mru enables unaligned access trapping on all his fate systems
[16:48:04] <mru> maybe this will get someone's attention
[16:48:14] <kshishkov> including x86?
[16:48:25] <mru> of course not
[16:48:44] <mru> only unaligned that can't be done in hw
[16:48:50] <kshishkov> well, it helped a bit with RV40 IIRC
[16:49:11] <mru> or put differently, I'm disabling unaligned fixup in kernel
[16:49:31] <kshishkov> that sounds more correct
[16:50:40] <mru> so soon there'll be lots and lots of yellow
[16:51:04] <Dark_Shikari> can you modify it so we can tell the difference between a failure and a crash?
[16:51:08] <Dark_Shikari> e.g. invalid output vs crash
[16:51:12] <Dark_Shikari> that would be rather nice
[16:51:30] <mru> open the details
[16:51:41] <mru> either the full report or click the arrow at the end of the line
[16:51:56] <mru> it says things like BUS or SEGV when it crashes
[16:52:31] <Dark_Shikari> Yeah, but it would be nice to sort
[16:52:37] <Dark_Shikari> i.e. to see which ones are crashes BEFORE clickong on each one
[16:52:48] <Dark_Shikari> so we don't have to manually search
[16:52:49] <av500> rss feed?
[16:52:52] <Dark_Shikari> I mean color coding
[16:52:59] <Dark_Shikari> say, yellow test ---> invalid output
[16:53:01] <Dark_Shikari> red ---> crash
[16:53:02] <mru> I don't understand what you want
[16:53:08] <Dark_Shikari> invalid output == yellow
[16:53:10] <Dark_Shikari> SIGX --> red
[16:53:11] <av500> mru: colours
[16:53:13] <mru> you mean if _any_ test crashes -> red
[16:53:14] <mru> ?
[16:53:16] <Dark_Shikari> no
[16:53:19] <Dark_Shikari> I mean in the LIST of failed tests
[16:53:36] <Dark_Shikari> Oh, it already has that.
[16:53:36] <kshishkov> colour-code return value for each test
[16:53:38] <CIA-93> ffmpeg: stefano * r24851 /trunk/ (6 files in 2 dirs):
[16:53:38] <CIA-93> ffmpeg: Rename av_fill_image_max_pixstep() to av_fill_image_max_pixsteps().
[16:53:38] <CIA-93> ffmpeg: The plural form is preferred as it is more consistent with the other functions:
[16:53:38] <CIA-93> ffmpeg: av_fill_image_linesizes()
[16:53:38] <CIA-93> ffmpeg: av_fill_image_pointers()
[16:53:38] <CIA-93> ffmpeg: and looks semantically more correct as it fills an array of elements.
[16:53:43] <Dark_Shikari> It's just not color coded
[16:53:45] <Dark_Shikari> I missed the column
[16:53:46] <mru> yes
[16:53:59] <Dark_Shikari> looks good then
[16:56:32] <CIA-93> ffmpeg: stefano * r24852 /trunk/doc/APIchanges: Add APIchanges for av_fill_image_max_pixstep() rename of r24851.
[17:03:09] <mru> meanwhile, anyone care to take a quick look at the swscale issue?
[17:03:23] <mru> warning: your head will spin and your eyes will bleed
[17:03:53] * kshishkov values his eyes too much
[17:04:05] <mru> try having the code read to you
[17:05:35] <mru> maybe I should get evil and fire up the old sgi/irix machine
[17:05:54] <kshishkov> please do
[17:06:26] <mru> dual 200MHz R10k
[17:06:59] <mru> or maybe install tru64 on the other alpha
[17:25:22] <av500> --enable-gpl  --enable-libfaac --enable-libfaad
[17:25:27] <av500> does that work?
[17:26:03] <Dark_Shikari> no, you need nonfree
[17:26:26] <av500> well, seagate did not think so
[17:26:49] <av500> and -formats says both faad and faac present
[17:26:59] <Dark_Shikari> old version of ffmpeg
[17:27:02] <Dark_Shikari> prior to the nonfree requirement
[17:27:04] <av500> 0.5
[17:30:20] <peloverde> Just because it wasn't blocked by configure doesn't make it legit
[17:30:36] <av500> thats what i think
[17:32:51] <av500> the supported codec list on that seagate dockstar aka pogoplug is impressive
[17:33:13] <peloverde> that said, when 0.5 was released people thought it was legit, so it's easy to claim ignorance
[17:33:52] * peloverde blames menno since it's exactly the same as the issue lame used to have and surely he was aware of that
[17:36:34] <kshishkov> av500: anything not supported by FFmpeg?
[17:38:24] <Dark_Shikari> mru: which systems are unaligned trapping enabled on?
[17:38:31] <av500> kshishkov: ?
[17:38:38] <kshishkov> alpha and probably mips
[17:38:43] <Dark_Shikari> not ppc?
[17:39:02] <Dark_Shikari> lavfi-pixfmts_scale_le failed with a sigbus, hah
[17:39:13] <kshishkov> av500: you said pogoplug codec list was impressive, so I ask if it will impress me
[17:39:20] <av500> no
[17:39:57] <av500> it will impress people in expensive suits...
[17:46:27] <peloverde> It looks like they are using r15625
[17:47:11] <kshishkov> av500: BTW, are you hinting we should sue them before anybody else does?
[17:47:27] <av500> yeah, get the little money they have
[17:48:40] <kshishkov> well, you can get payment in their harddrives
[17:48:57] <kshishkov> so we can get 10TB incoming filedump
[17:49:44] <av500> kshishkov: so, it will just fill up quickly, regardless of the size....
[17:50:03] <kshishkov> it will, but it will save us some time
[17:50:16] <av500> nah, ppl will just upload larger files
[17:50:26] <kshishkov> just forbid uploading MPEG streams or anime
[17:50:29] <av500> like full 4k 3D greenray rips
[17:50:46] <av500> with the smell track
[17:51:21] <kshishkov> that will be fun to debug
[18:20:20] <mru> Dark_Shikari: alpha, mips, and armv5 have full trapping
[18:20:40] <mru> and mips64 of course
[18:20:57] <kshishkov> and Chinese MIPS
[18:21:12] <mru> yeah, calling it mips64 is actually incorrect
[18:21:16] <mru> but it's close enough
[18:21:40] <kshishkov> there's official MIPS64 anyway
[18:21:44] <mru> yes
[18:21:49] <mru> it differs slightly
[18:22:12] <mru> there's mips64 v2 (or r2) as well
[18:22:19] <kshishkov> nobody cares about it, MIPS-III seems to be good enough for majority
[18:22:30] <mru> it has some very nice additions
[18:22:44] <mru> like mul with register destination
[18:22:51] <mru> lets you do more than one mul ever 5 cycles
[18:26:51] <mru> curious that dv fails with gcc 4.4 on alpha
[18:26:56] <mru> that's worth looking into
[18:27:17] <kshishkov> it would be more curious if it failed with suncc on solaris
[18:29:37] <CIA-93> ffmpeg: fenrir * r24853 /trunk/libavcodec/mpegvideo.c: (log message trimmed)
[18:29:38] <CIA-93> ffmpeg: Fixed mpeg12 top field first flag value with field picture encoding.
[18:29:38] <CIA-93> ffmpeg: The relevent extract of the iso 13818-2 about the value of the syntaxical
[18:29:38] <CIA-93> ffmpeg: element top_field_first of the Picture Coding Extension is:
[18:29:38] <CIA-93> ffmpeg: "top_field_first -- The meaning of this element depends upon picture_structure,
[18:29:38] <CIA-93> ffmpeg: progressive_sequence and repeat_first_field.
[18:29:38] <CIA-93> ffmpeg: [...]
[19:39:48] <BBB> I think I'm gonna give up on this mc setting, I can't get it faster than original code no matter what I try :(
[19:40:26] <Dark_Shikari> BBB: that's what I thought
[19:40:27] <BBB> Dark_Shikari: how much faster was your coreavc0D/ 1D/2D MC compared to always-2D?
[19:40:34] <Dark_Shikari> a good bit
[19:40:42] <Dark_Shikari> here's how it worked
[19:40:45] <BBB> so why can't I get it faster?
[19:40:49] <Dark_Shikari> if(!mx) {do 1D}
[19:40:52] <Dark_Shikari> if(!my) {do 1D}
[19:40:57] <Dark_Shikari> in the !mx, it check if(!my) at the start.
[19:41:18] <BBB> right, so that's what we do for width=8 already
[19:41:23] <BBB> but not for width=4/width=2
[19:41:28] <BBB> probably because the gain is too small
[19:42:02] <BBB> so that was all in one function, is what you're saying?
[19:42:09] <Dark_Shikari> I'm not sure about width4
[19:42:10] <BBB> (ather than having several smaller functions for each case)
[19:42:26] <BBB> I can test, that's not the problem
[19:42:41] <BBB> I'm just wondering if I'm using the wrong approach or if my result makes sense
[19:43:01] <BBB> and clearly having many small functions is slower than having a few bigger functions
[19:43:09] <BBB> as per my measurements
[19:50:36] <CIA-93> ffmpeg: mru * r24854 /trunk/libavcodec/dv.c: dv: fix alignment of scratch buffer
[19:52:39] * mru decides against waiting for dv maintainer review
[19:53:33] <kshishkov> "DV maintainer drunk"?
[19:53:51] <mru> he hasn't been heard from in a long time
[19:54:18] <kshishkov> year-long binges are not that rare in RUssia yet I agree with you
[19:56:56] <mru> he didn't live in russia last time I saw him
[19:57:11] <kshishkov> I know
[19:57:37] <kshishkov> but who knows where he is now and what he does, his company is nonexistent
[19:57:40] <mru> what's worse, he was at the time going for a full month without drinking due to losing a bet
[20:09:08] <Compn> hmm
[20:09:17] <Compn> wasnt there a patch posted to implement some 10bit codec ?
[20:09:19] * Compn cant find it
[20:11:58] <kierank> r10k you mean
[20:12:20] <Compn> yes, thats it
[20:12:32] * Compn kept searching for v210 ;
[20:13:05] <pJok> 10bit is out soon anyways
[20:13:23] <pJok> i heard 16bit float will be the new thing...
[20:29:53] <BBB> Dark_Shikari: no, it's slower also :(
[20:30:11] <BBB> either that or my asm skills are very crappy
[20:30:28] <mru> or both
[20:32:56] <BBB> quite possibly
[22:50:24] <saintd3v> peloverde: did you see the patches i sent to the ML?
[22:50:31] <peloverde> yes
[23:06:02] <saintd3v> ok, cool. i know you said it would be a little before you could review them. just wanted to make sure you didn't miss them.
[23:09:46] <Dark_Shikari> 08:34 < blight_> i thought av_resample functions would do resampling without changing pitch...?
[23:09:49] <Dark_Shikari> ]
[23:09:49] <Dark_Shikari> 09:06 < PovAddict> blight_: unfortunately I think you need to call av_wishful_thinking() first for that to work
[23:17:42] <CIA-93> ffmpeg: stefano * r24855 /trunk/doc/filters.texi: Add AUDIO FILTERS section.
[23:17:43] <CIA-93> ffmpeg: stefano * r24856 /trunk/ (5 files in 2 dirs):
[23:17:43] <CIA-93> ffmpeg: Add null audio filter.
[23:17:43] <CIA-93> ffmpeg: Patch by S.N. Hemanth Meenakshisundaram -af smeenaks,ucsd,edu.