Previous day | Jump to hour: 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 | Next day

Seconds: Show Hide | Joins: Show Hide | View raw
Font: Serif Sans-Serif Monospace | Size: Small Medium Large

Click in the nick column to highlight everything a person has said.
The Logo icon identifies that the person is a core developer (has commit access).

#rockbox log for 2005-10-26

00:00:37preglowman, i love overflow errors
00:04:32 Quit linuxstb (Read error: 110 (Connection timed out))
00:05:17 Join linuxstb [0] (
00:09:19 Quit ender` (Read error: 113 (No route to host))
00:13:37*linuxstb has just found the ffmpeg 24-bit FLAC bug
00:13:43linuxstbOverflow errors...
00:15:44 Join miner49er [0] (
00:15:56 Quit DrMoos (Read error: 104 (Connection reset by peer))
00:17:23preglowso it's a bug, not a feature?
00:18:16linuxstbThe problem is that ffmpeg itself is limited to 16-bit audio. Or at least, the flac decoder only outputs 16-bit samples at the end of the process.
00:18:34linuxstbSo nobody had tested 24-bit files.
00:19:25preglowhow well thought out
00:19:34linuxstbDoes your EMAC lpc_restore function deal with the "wide" case?
00:19:46linuxstbI was about to look, but as you are here...
00:19:49preglowi love this, the lowshelf filter goes on for about a second, then suddenly, starts bursting noise at certain intervals
00:19:54preglowlinuxstb: yes, i think so
00:20:08linuxstbCan you remember how much of a speed improvement those functions gave?
00:20:14preglowlinuxstb: not at all
00:21:30preglowbut yeah, libflac has lpc_restore routines for both 16 bits, and everything above, it seems
00:21:36preglowmy routine handles both right now
00:21:47preglowthat's one of the things i might optimise
00:22:11linuxstbYes - it has routines optimised for 32-bit ints (where overflow is not a problem), and 64-bit routines otherwise (mainly for 24-bit audio I think).
00:22:18preglowmaking a separate routine for 16 bit data, i think the 16 bit mac operation is faster
00:23:02linuxstbThere's an if statement in libFLAC that decides whether there is a risk of overflow or not. That was missing in ffmpeg, and ffmpeg always used the 32-bit version.
00:23:12linuxstbHence the overflow.
00:23:27linuxstbBut it was interesting to compare the ffmpeg and libFLAC code.
00:24:29preglowanything notable?
00:26:02linuxstbThe actual libFLAC decoding code is perfectly normal - it's the layer upon layer of code above it that's the problem.
00:26:35linuxstbIn fact, I'm tempted to change the ffmpeg decoder so it looks more similar to libFLAC - to try and spot any other possible bugs.
00:27:12preglowwhat i dislike about the libflac code is the structure
00:27:24linuxstbMaybe we were just using libFLAC in the wrong way - i.e. we should be using the lower-level functions and strip out a lot of the higher-level stuff.
00:27:34preglowsure, if that's possible
00:27:41preglowi haven't had a look at the api at all
00:28:23linuxstbBut I think ffmpeg is the way to go. I'm now getting matching md5sums for both 16-bit and 24-bit files, so it's good enough for me.
00:28:39linuxstbffmpeg just needs a little more tidying up.
00:30:04preglowany easily optimisable spots?
00:30:51linuxstbI haven't really looked at that side of it yet. The lpc_restore function is the obvious choice - using your EMAC routine from libFLAC
00:31:17linuxstbThe bitreading functions might be able to be optimised as well, but I haven't looked at those at all.
00:31:40preglowcould you paste me a source url again, please?
00:31:48linuxstbWe've also got about 10KB left of IRAM.
00:31:55linuxstbYes, I'll make a fresh archive.
00:33:01 Nick Gibbed is now known as Rick (
00:33:21preglowgod, i'm looking forward to when people will start asking for eq visualisation
00:34:32preglowyou can't use the libflac asm routines as is
00:34:58linuxstbWhy not?
00:35:44preglowlooks like arrays are set up slightly differently
00:35:57linuxstbI was just about to look at it before I gave you the source.
00:36:18preglowresiduals are stored in the sample arrays, for one
00:38:34preglowit's got a loop where i need to use fractional emac mode here
00:38:46preglowi wonder why i didn't have to do that in libflac
00:39:43linuxstbWhich loop is that?
00:40:07preglowforget it
00:40:13preglowi haven't optimised that for libflac either
00:40:35preglowdecode_subframe_lpc, lower part of the lower if
00:40:51linuxstbThe wsum part?
00:41:24preglowis that for files with a bitdepth of more than 32, or what?
00:41:27linuxstbThat's the code I just had to add to fix 24-bit output.
00:41:48linuxstbIt's the equivalent of lpc_restore_wide in libFLAC
00:41:55preglowyeah, i figured
00:42:09preglowwell, optimising that might prove hard
00:42:12linuxstbI haven't tested that on the iRiver yet.
00:42:27preglowwe might have to go the same route we did in libmusepack
00:43:02linuxstbI don't think it will be executed for 16-bit FLAC files though. So it's not that important.
00:43:07preglownor do i
00:43:31linuxstbAlso, FLAC is very bad at compressing 24-bit files. wavpack is a much better choice.
00:43:46preglowwavpack is a much better choice overall
00:43:49preglowif you ask me, at least
00:44:34linuxstbMy 24-bit/44.1KHz FLAC examples are almost 1500kbit/s
00:44:54preglowdamn, that's bad
00:45:03preglowhow much for wavpack?
00:45:10linuxstbI'll test now.
00:45:23preglowwavpack can even compress 32 bit floats
00:45:26preglowit's pretty l33t
00:45:48linuxstbThe file is 66% of the size - so I guess around 1000kbit/s
00:46:02linuxstbCompressed in 3 seconds.
00:46:26linuxstbOn a 2.8GHz Xeon.
00:46:56preglowbut right now i'd give a pretty penny to understand what the flaming hell is wrong with this filtering routine
00:47:20preglowcoef calculation is very precise, but the filtering itself seems to have some internal overflow problem
00:47:27preglowonly problem is it's supposed to be impossible...
00:47:29linuxstbSorry, I made a mistake. Wavpack is about the same as FLAC - 1432kps
00:47:48linuxstbThe FLAC is 1496kbps
00:47:56preglowahh, ok
00:48:03preglowabout the same margins as for 16 bit audio, then
00:48:18preglowwavpack tends to be slightly better
00:49:20linuxstbAny idea if wavpack is easier to decode than FLAC, or is it just that David did a very good job.
00:50:30preglowi have no idea, really
00:50:33preglowi think a bit of both
00:52:05linuxstbI'm hoping we can get FLAC to a similar level.
00:53:21preglowwould be nice, yeah
00:55:52linuxstbOuch. My 24-bit files have about a 70% boost ratio. Compared to less than 10% for 16-bit files.
00:56:24linuxstbBut they are playing perfectly.
00:56:41preglowwell, go figure
00:56:51preglowit's suddenly 64 bit mul-o-rama in that code path
00:57:15preglowrest assured we can slice that drastically
00:57:20preglowbut it'll always be slower, of course
00:57:34linuxstbI hope so. It's now boosting near 100% and pausing.
00:57:48preglowwe can use the same trick we did in libmusepack
00:58:36*linuxstb goes and checks the cvs logs
01:00:24preglowcombining an emac mul and an ordinary mul to do a 64 bit mul
01:00:40preglowamiconn also codec a much faster 64 bit shift
01:02:12linuxstbDo you think we will get lots of complaints if I commit this new decoder before adding seeking support?
01:02:43linuxstbIt seems that seeking is actually quite hard in FLAC files - mainly because each compressed frame doesn't include the frame length.
01:03:43preglowbut flac files have seek tables, yeah?
01:04:12preglowdon't know about complaints, but i'd say just commit it
01:04:17linuxstbYes, but I think you still need to walk through the stream a little afterwards.
01:05:01linuxstbHave you read the bug reports on the forum thread I started about this decoder?
01:05:38linuxstbI think these are general problems, not specific to my new decoder. But I'm not sure.
01:05:55linuxstbIt seems some people are using old installations as well.
01:06:06preglowi'm sure they are as well
01:06:19preglowcan't imagine how it can be otherwise
01:07:48linuxstbOK, I'll probably commit it tomorrow. Just to give others a chance to complain if they read this in the logs.
01:10:31 Quit miner49er ("CGI:IRC")
01:11:26preglowsounds good
01:15:19*preglow kicks the lowshelf filter
01:15:21***Saving seen data "./dancer.seen"
01:15:46linuxstbAlso, does anyone know why plugins/Makefile still links everything against the codec libraries? Is this just a leftover from the xxx2wav plugins and can be deleted?
01:17:48preglowno idea
01:17:52 Quit SeeSchloss (Read error: 110 (Connection timed out))
01:27:34linuxstbLooks like podzilla now runs on the Nano - so IPL has arrived.
01:34:40preglownow there's a good excuse to get one if i ever heard one
01:35:09linuxstbI'm just about to test podzilla for the first time on my ipod.
01:37:00amiconnpreglow: As for the 64 bit shift-by-n routine, the idea is taken from the gcc routine, of course optimised ;)
01:38:44amiconnIt also profits from the fact that we only need 32 bits of the final result
01:42:47preglowok, i'm really starting to hate fixed point now
01:43:02amiconnFixed point is great ;)
01:43:14amiconnmandelbrot.rock uses fixed point as well
01:43:39 Quit cYmen ("zZz")
01:43:51amiconnFormat is 6.26, or 6.10 for low precision
01:44:02preglowyes, it's nice enough, but i'm not used to these kind of errors!
01:46:25preglowi'm really starting to grow sceptical about how fast a five band eq will be
01:48:13preglowquite luckily the filtering loop should be a breeze to write in asm
02:00:08TiMiDanyone here still awake who speaks english well ?
02:09:34linuxstbI'm almost awake and English is my first language.
02:11:41TiMiDcool ;)
02:12:01TiMiDwell I wrote a motivation letter
02:12:15TiMiDto find an internship in japan
02:12:43TiMiDbut since english is not my mother tongue, it may not be very good
02:12:53TiMiDwould you mind looking at it ?
02:13:23linuxstbI was about to go to bed. But if it can wait about 8 hours, I'll be happy to look at it in the morning.
02:13:55linuxstbOh go on then. dave at :)
02:14:05TiMiDI would like to send it right now for them to read it in the morning
02:14:23linuxstbYep - send it now and I'll have a look.
02:14:56TiMiDi put it on a ftp
02:15:01linuxstbno problem.
02:15:10TiMiDjust tell me the big fault, it hasn't to be 100% correct
02:16:04linuxstbIt's already 9.15am in Tokyo...
02:16:28TiMiDyes I know :)
02:16:34TiMiDI'm late :/
02:16:37 Join tvelocity [0] (
02:16:47TiMiD(it's not very long)
02:17:42linuxstbs/finishing in my/finishing my/
02:18:06linuxstbProbably best to write "I am" instead of "I'm" in a formal letter.
02:18:33TiMiDok I remve all the ' :p
02:18:56linuxstbingineer should be engineer
02:19:37TiMiDouch I'm not very awake too :)
02:19:45linuxstbIn fact, I'll just edit it slightly and upload a new version.
02:20:18TiMiDthnk you very much (you save my life !)
02:22:26preglowjapanese people kill people with bad english? :P
02:22:48preglowno wonder they've got plenty of suicides, then
02:23:01amiconnlinuxstb: I just had a quick look at alac.c - your clz routine is indeed rather inefficient
02:23:24amiconnDunno how much it is called though
02:23:40TiMiDpreglow: hmm I don't even know if they understand english well :p
02:23:59amiconngcc has a rather clever clz routine
02:24:24preglowTiMiD: most people don't
02:24:52preglowso you'll have plenty of incentive to brush up on your spoken japanese
02:25:11linuxstbJust minor changes - it was basically fine.
02:25:32TiMiDthank you very very much :)
02:25:34linuxstbI have one problem - "passionated" isn't a word. But I can't think of an alternative in that context.
02:26:03 Join ashridah [0] (
02:26:24linuxstbamiconn: What is the clz routine?
02:26:58TiMiDhow would you say to have a passion ?
02:27:20linuxstbThat's the problem, I can't think of a way to say that. I guess English people don't have passions :)
02:27:31TiMiDhehe :)
02:28:11TiMiDI don't see another way to say it either :/
02:28:17linuxstbamiconn: Yes. But it's not my function. The author even admits how awful it is.
02:28:24TiMiD(in french you would say it like that)
02:28:38amiconnI could find 'passionate' in my English-German dictionary
02:28:47preglowdoesn't fit
02:29:07linuxstbYes - just say "I am passionate about ..."
02:29:36amiconnlinuxstb: Okay, so it's by the alac author, and yes, I read the comment
02:29:38TiMiDimpassioned works ?
02:29:47TiMiD(I found it in the dictionnary)
02:30:07linuxstbTiMiD: "I am passionate about" fits your sentence perfectly.
02:30:13TiMiDok :p
02:30:57amiconnlinuxstb: I hope you used the 'correct' iram attributes for your new flac codec? ;)
02:31:28linuxstbDo they work in the codecs?
02:31:28preglow Zero Count
02:31:41amiconnlinuxstb: yes
02:32:22TiMiDwhen I think they will more probably use an automatic translator to understand this beautifull :(
02:32:25amiconnThere's no real difference yet, except ICONST_ATTR allows putting const data into iram without dropping the 'const'
02:33:05amiconnUsing the correct attributes will help to see whether it's worth adding proper .ibss handling for codecs
02:33:23TiMiDhey let him go to bed now :)
02:33:31linuxstbI think it will be - flac.codec for example reduces down to about 8300 bytes without the buffers.
02:33:52amiconnNice :)
02:34:02linuxstbIt's scary.
02:34:21preglowso there's room for all of the codec in iram?
02:34:29linuxstbpreglow: Yes, that's what I said the other day.
02:34:51TiMiDhow much is your codec in size ?
02:34:53preglowthere's still a lot of unused iram around for all codecs
02:35:00linuxstbTheres about 38KB of buffer and 8K of code.
02:35:11TiMiDhehe :)
02:35:22linuxstbPesky clipboard. Never forgets.
02:35:28TiMiDnono :)
02:35:35TiMiDit's a virus in my .txt
02:35:46TiMiDexploits a buffer averflow in notepad :p
02:36:25linuxstbBack to the IRAM - I should be using IDATA_ATTR for the data, and ICONST_ATTR for any lookup tables?
02:37:27amiconnIDATA_ATTR for initialised data, ICONST_ATTR for constant data and (most important) IBSS_ATTR for uninitialised data, e.g. buffers
02:37:56amiconnI think IDATA_ATTR will actually be used least in codecs, if at all
02:38:23linuxstbYes - there is no use of it in my new FLAC decoder now.
02:41:08amiconnEven the core uses very little idata - 8 bytes (!) (h120)
02:41:22amiconnPlus 16 bytes iconst
02:41:27 Quit tvelocity ("Leaving")
02:43:42 Join webguest82 [0] (
02:47:32 Quit gromit` (Remote closed the connection)
02:47:46TiMiDhi webguest82
02:47:55preglowthere already is a db lookup table some place in rockbox?
02:50:26TiMiD28 mails to send
02:50:41TiMiDthe night will be short
02:53:04TiMiDbye ?
02:53:14TiMiDyou are short
02:53:34webguest82i live in korea
02:53:39*preglow slaps the lowshelf filter
02:53:40TiMiDI know :p
02:53:45preglowi'm off to bed, night all
02:53:47TiMiDI read the logs
02:54:24webguest82It goes to the school.
02:54:43webguest82good night
02:55:00 Quit webguest82 ("CGI:IRC (EOF)")
03:02:46linuxstbGoodnight all. Good luck TiMiD.
03:05:16 Join gromit` [0] (n=gromit`
03:06:47TiMiDthank you :)
03:10:38 Quit lostlogic (Read error: 110 (Connection timed out))
03:15:23***Saving seen data "./dancer.seen"
03:33:43 Join Midgey34 [0] (
03:41:55 Join lostlogic [0] (
03:44:52 Quit gromit` (Remote closed the connection)
03:59:58 Quit Midgey34 ("CGI:IRC")
04:04:14 Join alxcm [0] (
04:04:18alxcmsup ppl?
04:04:47 Join Midgey34 [0] (
04:06:09 Join _Vladoman [0] (
04:08:52alxcm"Codec Failure"
04:09:03alxcmwhen trying to play anything using today's daily build on my ihp120
04:17:18 Quit QT (Read error: 110 (Connection timed out))
04:21:36 Quit novimon (Read error: 110 (Connection timed out))
04:23:59 Join gursikh [0] (
04:24:20 Quit Vlad0man (Read error: 110 (Connection timed out))
04:39:48 Part gursikh
04:56:01 Join myndzi [0] (
04:56:08 Part myndzi
05:15:26***Saving seen data "./dancer.seen"
05:22:05 Nick Midgey34 is now known as MijZZZ (
05:27:09 Join webguest56 [0] (
05:28:45 Quit webguest56 (Client Quit)
05:45:08 Nick ashridah is now known as Lost-ash (
06:02:21 Join MijZZZ_ [0] (
06:07:43 Join MacDancer [0] (
06:18:20 Quit MijZZZ (Read error: 110 (Connection timed out))
06:24:04 Nick Lost-ash is now known as ashridah (
07:01:46 Join amiconn_ [0] (
07:15:27***Saving seen data "./dancer.seen"
07:17:07 Join matsl [0] (
07:17:57 Join SeeSchloss [0] (
07:20:07 Quit amiconn (Read error: 110 (Connection timed out))
07:20:08 Nick amiconn_ is now known as amiconn (
07:53:45 Quit ghode|afk (Read error: 110 (Connection timed out))
08:02:46 Quit MacDancer ("Leaving")
08:05:21 Join ender` [0] (i=ychat@
08:11:45 Join B4gder [0] (
08:11:47 Quit matsl (Remote closed the connection)
08:20:48 Join Lost-ash [0] (
08:21:10 Quit ashridah (Nick collision from services.)
08:23:25 Nick Lost-ash is now known as ashridah (
09:03:40 Join einhirn [0] (
09:06:59 Part amiconn
09:12:06 Join solexx [0] (
09:12:53 Join amiconn [0] (
09:15:29***Saving seen data "./dancer.seen"
09:27:11 Quit solexx_ (Read error: 110 (Connection timed out))
09:27:37 Quit Lynx_awy (Read error: 104 (Connection reset by peer))
09:36:34 Join Lynx_ [0] (
09:41:28phaedrus961amiconn (or anyone who's interested): I've updated the unicode patch with the changes you asked for
09:41:41phaedrus961could you have another look when you have time?
09:42:19phaedrus961I also managed to shave a couple hundred bytes off the binary size :)
10:06:42 Join Moos [0] (
11:05:07 Join cYmen [0] (
11:15:30***Saving seen data "./dancer.seen"
11:26:10amiconnphaedrus961: yes, tonight
11:26:53 Quit cYmen (Read error: 104 (Connection reset by peer))
11:29:48 Quit markun (Read error: 104 (Connection reset by peer))
11:30:33 Join muesli- [0] (
11:30:42 Join cYmen [0] (
11:30:59 Join markun [0] (
11:33:12 Join DangerousDan [0] (
11:35:53 Join cYmen_ [0] (
11:41:07 Join cYmen__ [0] (
11:48:42 Quit cYmen (Connection timed out)
11:49:10linuxstbMorning all.
11:50:33markunGood mornig
11:50:46linuxstbpreglow and I are both in favour of comitting the new flac decoder immediately (even though there is no seeking yet). Does anyone object?
11:50:51muesli-g'day m8s
11:51:20markunFirst step to arabic rockbox:
11:52:48markunthe first line shows the current rendering. The second line is produced with a script that makes it joining.
11:53:43 Quit cYmen_ (Connection timed out)
11:55:13 Join linuxstb_ [0] (
12:00:45linuxstb_I've just realised why plugins/Makefile still links against the codec libs - the wav2wv plugin.
12:03:53 Quit linuxstb (Read error: 113 (No route to host))
12:17:55 Quit linuxstb_ (Read error: 104 (Connection reset by peer))
12:21:40 Join linuxstb_ [0] (
12:24:22 Quit MijZZZ_ ("Chatzilla [Firefox 1.0.7/20051010]")
12:26:12 Nick ashridah is now known as Lost-tv (
12:27:58 Quit muesli- (Read error: 113 (No route to host))
12:31:01 Nick linuxstb_ is now known as linuxstb (
12:52:03linuxstbStrange. I was playing music, then plugged in USB (with the music playing), and now I can hear a very faint radio signal in my headphones.
12:52:17linuxstbI haven't listened to the radio for days.
12:53:08SlasheriHmm, that must be some other interference :)
12:53:40linuxstbMaybe. I just left USB mode (back to file browser), and it's silent again.
13:06:03 Quit ryan_j (Read error: 110 (Connection timed out))
13:15:34***Saving seen data "./dancer.seen"
13:20:02 Join ryan_j [0] (
13:21:22 Join ripnetuk [0] (
13:27:53 Join _DangerousDan [0] (
13:33:14 Nick Lost-tv is now known as ashridah (
13:35:38 Quit ripnetuk ("Ninja IRC v1.5.8.1(#1) exiting after 14m16s of use")
13:40:31 Join Vlad0man [0] (
13:43:23 Quit DangerousDan (Read error: 110 (Connection timed out))
13:58:15 Quit _Vladoman (Read error: 110 (Connection timed out))
14:01:12 Join muesli- [0] (
14:11:43markunhi muesli-
14:37:17B4gdernice commit linuxstb!
14:37:33*preglow applaudes
14:37:40*linuxstb bows
14:37:59linuxstbNow the fun commit - removing libFLAC :)
14:38:21preglowcan't that wait for a bit?
14:38:31preglowwe might for some reason need to put it back in again
14:38:50preglowmight as well wait until the new flac decoder is fully operations
14:38:54preglowoperational <-
14:39:05linuxstbOK. I removed it from FILES, and it's not compiled any more. So there's no harm in leaving it there.
14:39:45linuxstbpreglow, amiconn: Please feel free to work some EMAC magic on it.
14:41:39preglowwill see what i can do
14:42:01preglowthink i'll just modify the unrolled libflac one and drop it in somehow
14:46:53 Join novimon [0] (
14:48:46preglowbtw, does the wavpack decoder play 24 bit files with a big speed penalty as well?
14:48:51preglowif i remember correctly, it shouldn't
14:49:56linuxstbI remember seeing lots of 16-bit/44.1KHz assumptions in the code, but I haven't checked recently.
14:50:17linuxstbMaybe that was just the old wv2wav plugin though.
14:51:00preglowi think so
14:51:09preglowwavpack already uses a 64 bit mul, even for 16 bit data
14:51:14preglowso i think it should be able to cope very nicely
14:51:16novimonw00t, new flac encoder!
14:51:19novimonmust check out
14:51:24preglowdecoder, even
14:51:25novimondecoder! :D
14:51:31novimonthe old one was bugged
14:51:48preglowlinuxstb: you think it'll be hard to add seeking?
14:51:49linuxstbHow do you mean bugged? The only problem I knew about was that it was slow.
14:52:00linuxstbpreglow: Not really. Just needs a couple of hours of concentration.
14:52:16linuxstbI should be able to do it in the next few days.
14:52:21novimonyeah the seeking was very very slow and the decoder used to skip
14:52:29novimonnot so often thou but from time to time
14:52:52linuxstbSadly, seeking is quite hard in raw FLAC files. OggFLAC would be better in this regard.
14:53:39linuxstbThe size (in bytes) of a compressed frame isn't stored in the frame header. So you need to partially decode the frame in order to find its length.
14:54:48linuxstbI'm not sure how accurate the seek table normally is, but I'm guessing the implementation will be to seek to the closest frame using the seek table, and then walk the stream until we get to the correct frame.
14:55:58preglowa non-accurate seek table
14:56:00preglowwhat a concept
14:56:18linuxstbI mean the resolution of the seek table.
14:58:15preglowahh, like that
14:58:45linuxstbJust checked - the default behaviour of the reference encoder is to insert seek points every 10 seconds.
14:58:51 Join webguest82 [0] (
14:59:35preglowlinuxstb: that's not bad
15:00:02linuxstbNo. I'll probably do a first implementation that just seeks to the nearest seek point.
15:00:32linuxstbIt will be better than nothing.
15:00:55amiconnWhy no interpolate between seekpoints, like with mp3 vbr?
15:01:22preglowthat'll come
15:01:51preglowit'll become important when we are to support cue files
15:02:16linuxstbThat's one approach - but you then have to sync to the start of the frame. Which is not easy as the sync code is not guaranteed to not appear inside frames.
15:02:30amiconnImho we shouldn't support .cue files at all
15:02:41preglowwhy not?
15:02:51novimon.cue support would be awesome
15:03:01linuxstbIf a FLAC file has an embedded cue file, then the cue points should have corresponding entries in the seek table.
15:03:22amiconn.cue files are a broken concept, imho
15:03:35amiconnThey are for when the player doesn't support gapless.
15:03:39preglownevertheless there are a lot of files using them out there
15:04:07linuxstbThey have other uses - think of them as "chapter points" in a long recording. Other formats (such as m4a) include chapter points.
15:04:24linuxstbIt is nice in some situations to not have to split your long file into individual tracks.
15:05:04linuxstbBut I agree that they are often used as a workaround for players not supporting gapless.
15:05:23preglowsure, me too
15:05:31preglowbut there are still nevertheless quite a lot of them around
15:05:39preglowand i'd like to have them supported
15:05:44preglowwavpack files also support embedded cue files
15:05:48preglowflac too, i believe
15:06:18linuxstbYep. People use them to backup CDs to a single file.
15:11:07amiconnReally? I can't believe that, especially with lossless
15:12:03linuxstbI don't really understand, but I think a cuefile is needed for a 100% accurate backup. Storing things like CD-TEXT, hidden tracks etc
15:12:06preglowwhat's so hard to believe about it? i'd do it myself if i ever backed up my cds losslessly, heh
15:15:37***Saving seen data "./dancer.seen"
15:15:49 Join webguest38 [0] (
15:15:59amiconnThere are 2 points I don't understand: (1) Why should I backup a CD I have myself as lossless? (2) If I had a reason, why would I convert it to a single file and cuesheet?
15:16:31linuxstb1) The same reason you back up anything of value
15:16:34amiconnI'd rather use individual files per track. They are more convenient to handle, especially with lossless
15:16:45amiconnLossless files are huge
15:16:48webguest82hi webguest38
15:17:06 Quit webguest38 (Client Quit)
15:18:11linuxstb2) Because you want to preserve the CD perfectly. But on this point I agree with you - backing up to individual tracks is good enough for me.
15:19:27 Quit Zagor (Read error: 110 (Connection timed out))
15:20:32amiconnlinuxstb: I consider the CDs as my backup. I only rarely play directly from CD once I ripped them (to lossy format, single tracks)
15:22:14 Join Zagor [0] (n=bjst@pdpc/supporter/sustaining/Zagor)
15:22:43linuxstbI'm not saying that _you_ must back up your CDs. Just that lots of other people do. I'm guessing those people actually play their CDs in a CD player.
15:22:59ashridahheh. i rarely play cds directly.
15:23:06ashridah(for obvious reasons)
15:23:28preglowi play cds all the time when i've got my cd player rigged up
15:23:30preglowwhich i haven't now
15:23:31linuxstbThere is always the risk of theft as well.
15:23:45preglowbut i understand why people back up cds perfectly
15:23:50preglowi just don't do it myself
15:24:13ashridahlinuxstb: that one's bitten me. lost a lot of cds i didn't have in mp3 format at the time (ogg wasn't even v1.0 back then, flac was unknown to me)
15:25:36 Quit ashridah ("Leaving")
15:28:50 Quit novimon ("Changing server")
15:35:34 Quit webguest82 ("CGI:IRC (EOF)")
16:09:41 Quit SeeSchloss (Read error: 110 (Connection timed out))
16:29:53 Quit B4gder ("time to say moo")
16:50:05 Join _FireFly_ [0] (
16:59:38 Join actionshrimp [0] (
17:15:40***Saving seen data "./dancer.seen"
17:26:42 Quit muesli- (Read error: 113 (No route to host))
17:32:12 Join SeeSchloss [0] (
17:48:22 Join XavierGr [0] (
18:00:49 Join paugh [0] (n=kickback@2001:5c0:8fff:ffff:8000:0:3e03:6822)
18:10:54 Join ghode|afk [0] (
18:24:57 Join tvelocity [0] (
18:26:12 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
18:45:13 Quit Mxm`Pas`Bien (Read error: 104 (Connection reset by peer))
18:45:20 Join Maxime [0] (
19:06:04 Quit Maxime ()
19:15:18 Quit SeeSchloss (Read error: 110 (Connection timed out))
19:15:40 Join SeeSchloss [0] (
19:15:41***Saving seen data "./dancer.seen"
19:20:14 Nick paugh is now known as AliasCoffee (n=kickback@2001:5c0:8fff:ffff:8000:0:3e03:6822)
19:43:10 Nick Lynx_ is now known as Lynx_awy (
19:52:29 Quit Lynx_awy (Read error: 104 (Connection reset by peer))
19:52:49 Quit linuxstb (Read error: 104 (Connection reset by peer))
19:53:10 Join linuxstb [0] (
19:55:21 Join linuxstb_ [0] (
20:04:23 Quit linuxstb (Read error: 113 (No route to host))
20:05:57 Join dpassen1 [0] (n=dpassen1@resnet-233-61.resnet.UMBC.EDU)
20:19:32 Quit dpassen1 ()
20:38:07 Quit _FireFly_ ("Famous last words")
20:48:38 Join Febs [0] (
21:15:42***Saving seen data "./dancer.seen"
22:20:30MoosHey guys! very quite day today :-)
22:21:25Moosmarkun/phaedrus961/amiconn: what's new about unicode ?
22:23:19Moosor preglow: what's new about your EQ? :D
22:28:19CoCoLUSwhy was the flac codec replaced?
22:28:44XavierGrbecause the new one is much faster.
22:29:09XavierGrdecoding happens almost without cpu boosting.
22:29:20CoCoLUSah... nice :)
22:36:45TiMiDhi Moos
22:37:07TiMiDtoday 55 mails ^^
22:37:32Mooshi btw
22:40:17 Quit dwihno (Read error: 110 (Connection timed out))
22:44:09 Join Bagder [0] (
22:49:26XavierGr55 mails? where?
22:50:56TiMiDjapan companies
22:51:15preglowMoos: not much new, something wrong i need to figure out
22:51:39preglowonce that is done, i just need to integrate it into rockbox and make some gui for it
22:51:40Moosok, have a good "luck" :)
22:51:42XavierGrTiMiD: Why are you going to work over Japan?
22:51:54Moospreglow: oh, that sound good
22:52:25preglowMoos: might still take some time, don't hold your breath
22:52:25 Join dwihno [0] (n=dw@
22:52:29TiMiDXavierGr: I'm indeed trying to
22:52:35XavierGrpreglow: Where do you think that the user will get the EQ gui? Inside Playback settings maybe?
22:52:54Moospreglow: but good to hear you are working on it :)
22:53:08preglowXavierGr: well, i guess so
22:53:10XavierGrTiMiD: So you are going to move over to Japan? Why go this far?
22:53:26preglowXavierGr: personally i'd like a full graphical eq like screen, but i can't be bothered to do that
22:53:32XavierGrYou are in France, aren't you? You can choose another EU country as well
22:53:45TiMiDXavierGr: because I'm a manga maniac ^^
22:53:47Moosor the FRANCE as well :)
22:53:50preglowthere is quite a difference between an eu country and japan, you know
22:54:17TiMiDI don't know why
22:54:28XavierGrpreglow: That would be great, though once the EQ code is done the gui is just detail and can be done by abyone.
22:54:29TiMiDI'm attracted by this culture and language
22:54:46TiMiD(nothing to do with mangas though)
22:54:46preglowXavierGr: yes
22:54:52TiMiDit's just that
22:54:57preglowi'd love to learn japanese
22:55:07XavierGrSo that is why you took Japanese as a lesson in your college, right? And what about your studies, did you finished?
22:55:10preglowi tried once, but i stopped after a while
22:55:59Moosdepand in the motivation, if you want to work in Jap you could learn more easily
22:56:01XavierGrI am learning Japanese through Anime series and movies! :)
22:56:09TiMiDI will be done in january
22:56:13preglowhad no use for it apart from playing snes rpgs in their original language anyway, heh
22:56:13XavierGrKonni chioua!
22:56:18preglowXavierGr: ahahah
22:56:29preglowyou have some work to do on your spelling, young man!
22:56:42TiMiDXavierGr: learn the kanas first, that will help you :)
22:56:47preglowTiMiD: i did
22:56:55TiMiDKO N NI CHI WA
22:56:55preglowTiMiD: forgotten most of them now, though
22:57:08TiMiDor HA
22:57:22preglowthere's only one japanese word you need to know: BAKA
22:57:23TiMiDif you don't prictice them it's hard :)
22:57:32TiMiDI prefer AHOU
22:57:38XavierGranta baka?
22:58:03XavierGrthis is my favorite word! Especially wih the "anta" in front of it
22:59:19TiMiDanata no baka
22:59:24XavierGrSo in what sort of job are you sending you CV for?
22:59:34TiMiDfor computer engineer
22:59:37preglowbaka yo
22:59:55dwihnoHaiyaku, torihada-chen!
23:00:01XavierGrattashi baka!
23:00:12TiMiDwatashi :)
23:00:28XavierGrI alwasys forget that
23:00:36XavierGrbut I should have said boku
23:01:21TiMiDyou'r not a woomen (who knows ^^)
23:01:29XavierGrwatashi sounds more cute! :)
23:01:50XavierGrboku ~ baka you know....
23:02:30preglowlinuxstb_: my, what i nice little codec libffmpegFLAC is
23:03:02linuxstb_Yes, I know :)
23:03:05TiMiDif you say bak
23:03:14TiMiDthen tou can call yourself 'ore'
23:03:25TiMiDsince you don't use polite language ...
23:03:26XavierGrspeaking of it which is the most optimized codec so far?
23:03:28preglowlinuxstb_: but where does it store all the residuals and lpc and so on? does it use the pcm buffer for it or something?
23:03:41linuxstb_I've just tested wavpack with 24-bit files, and they play fine, about 25% boost.
23:04:16XavierGrTiMiD: so any luck to get a job until now?
23:04:18linuxstb_Yes, I think they are stored in the "decoded" arrays - these are the only buffers it uses.
23:04:26amiconnlinuxstb_: 36 KB ibss, ~670 bytes iconst, no idata
23:04:38amiconnI hope the other codecs are similar...
23:04:45preglowlinuxstb_: excellent, just excellent
23:04:49linuxstb_I removed the final "convert to 16-bit and copy to different buffers" stage from the codec, so it now does it everything in-place.
23:04:53preglowlinuxstb_: the libFLAC people have got a thing or two to learn here
23:05:02TiMiDXavierGr: time will tell but I'm not vry confident ...
23:05:12TiMiDI'll try USA if japan fails
23:05:28TiMiDat least I would have done the maximum I could do so no regret
23:05:38preglowXavierGr: the most optimised codec is probably libmad
23:05:50XavierGrTiMiD: Truth is that there must be a lot of competition there for this kind of job.
23:05:55preglowXavierGr: as in that's what's seen most optimisation work
23:06:10linuxstb_Silence all day, and then everyone talks at once....
23:06:49XavierGrpreglow: libmad is vorbis? or...
23:06:55preglowXavierGr: mp1/2/3
23:07:01amiconnIf we would add separate bss and ibss handling, flac.codec would be a mere 9 KB...
23:07:20XavierGrah yes the MAD
23:08:45preglowamiconn: and while i remember it, have you tested mp1 performance?
23:09:24amiconnI have just one mp1... but I could probably produce some test files with different bitrates
23:09:40preglownothing imporant, i'm just wondering how good/bad it is
23:09:45preglowi haven't got even one mp1 file
23:09:59amiconnI can create them, with Pegase on Amiga
23:10:16amiconnIt's a rather good layer 2 encoder, but also does layer 1
23:10:59amiconnIt's *really* fast compared to lame on Amiga, although not realtime
23:11:09 Quit _DangerousDan (Read error: 104 (Connection reset by peer))
23:11:17linuxstb_So far I've managed to avoid malloc entirely in the new flac codec (libffmpeg had some which I removed). But the seek table will vary in size from about 200 bytes (for a typical 4 minute song with standard 10 second seek points) up to maybe 40KB-50KB for a very large (i.e. 2GB) file.
23:11:44amiconnOn my A4000/060/50, lame is ~10% realtime with cbr and ~3% realtime with vbr. Pegase is ~50% realtime...
23:12:46amiconnToo bad it uses floating point...
23:15:38preglowi wonder when christian gmeiner will return
23:15:43***Saving seen data "./dancer.seen"
23:15:59 Join muesli- [0] (
23:16:00BagderI mailed him again a couple of days ago
23:16:19Bagderno reply so far
23:16:50preglowyou bought a bdm for him, no?
23:17:11BagderI asked him to mail it back to us if he doesn't plan on using it for Rockbox
23:17:28amiconnPerhaps I should do some layer 1 and layer 2 test files and put them on my webspace. It sits mostly unused...
23:17:46preglowamiconn: would of course be nice, i have few such files myself
23:17:57preglowBagder: yes, giving it to some other interested party would be nice
23:18:06amiconnpreglow: What do you think would be useful bitrates to test with? Should I use our famous digital flow thingy?
23:18:49preglowamiconn: you can do whatever you want in that regard, as for bitrates, the extremes would be nice, plus something in between
23:19:14preglowi think 44.1 khz will do for sample rate, i don't see any point in testing the resampler any further
23:19:24preglowi know it sucks anyway ;)
23:22:33amiconnI have 200MB webspace of which I'm currently using ~2MB only...
23:23:01TiMiDmay I ask what the current status of my patch is ?
23:23:02 Quit ender` (Read error: 104 (Connection reset by peer))
23:23:06preglowi've got five gigs of which i use around 10...
23:23:15preglowmegs, that is
23:23:26TiMiDI'm a little bored to update it each time someone updates a file I modify
23:23:50preglowTiMiD: afaik, none of the core devs have time to look at it
23:23:57preglowi've got no idea about that area of rockbox, so can't help
23:24:48preglowi think i'll see about adapting my libflac lpc patch
23:25:50markunamiconn: Do you have time to try out the unicode patch? There are some problems with the text viewer, but phaedrus961 said he will take a look at it.
23:25:53XavierGrpreglow do you have any realtion to the FM radio cado for iriver? I've got the preset file patch pending... :p
23:26:07preglowXavierGr: no relation, no
23:26:11TiMiDall patches coming at the same time ^^
23:26:16preglowcome to me for codecs and dsp, heh
23:26:54MoosTiMiD: the time when Linus, Bagder... apear don't have lot of time :-(
23:26:57TiMiDI only got LinusN and amiconn opinions about remote
23:27:10Mooswe all want it sure :)
23:27:10TiMiDother one have looked at it ?
23:27:15TiMiDdon't be shy :)
23:27:25preglowTiMiD: well, what more do you want?
23:27:44Mooscommit :D
23:27:54preglowTiMiD: if you've got linus' and amiconn's opinions, you've got some very influential opinions and should probably do what they say ;)
23:28:01XavierGrhow was the feedback of Linus and amiconn, bar or good?
23:28:07TiMiDI've done it :p
23:29:11amiconnpreglow: I'll do a 'typical' and a maximum bitrate version for layer 1 and 2 each
23:29:13TiMiDgood :)
23:29:19TiMiDI think
23:29:38amiconnlayer 1: 256/448 kbps layer 2: 192/384 kbps
23:29:40markunLinusN was quite positive I think
23:29:54Moosyeah he liked it :)
23:30:23Moosbut those last weeks he don't have time for Rockbox
23:30:56XavierGrI cant wait for commitment, imagine full remtote support and full recording support that Slasheri is working on.
23:31:09MoosBagder: any opinion?
23:31:29TiMiDit would be good then that some people test it on real targets (I only hve tested on sim and iriver)
23:31:52TiMiDnad then if it doesn't beaks anything why not a commit
23:33:07MoosTiMiD: why don't continu to work on it, currently just file tree; right?
23:33:31XavierGrwell I feel him
23:33:43TiMiDbecause it would be such a huge work to maintain it
23:33:45Moosyou already know it will be one day or other for sure
23:33:54XavierGrif this is not going to be commited there is no point to continue
23:34:00Moosit will
23:34:13TiMiDbut it's still not in it
23:34:14Moosbut when, that is the question :)
23:34:16markunTiMiD: Does it break a lot when you use cvs update?
23:34:27TiMiDbreak ?
23:34:42 Join DangerousDan [0] (
23:34:45markunOr do you not use cvs update?
23:34:55TiMiDI use CVS update
23:35:06TiMiDbut I can't do it on my current source tree
23:35:16markunWhy not?
23:35:22TiMiDsince some files are deeply modified
23:35:46amiconnI think as soon as either remote lcd or unicode gets committed, the other patch will suffer from a number of conflicts
23:36:15markunYes, probably.
23:36:42markunalthough I applied both TiMiD's patch and the unicode patch..
23:36:57TiMiDI don't mind if it's one other big update
23:37:03markunCan't remember many conflicts
23:37:21preglowi love big changes
23:37:27TiMiDsince I understand the code I can modify it
23:37:53TiMiDbut what I don'tlike is contant little changes to files day by day :D
23:38:06TiMiD(welle it happens twice a week)
23:39:21XavierGrTiMiD: I know exactly what you mean. I had to that for the remote patch (before Firefly). Though I knew that this kind of buggy work would never be commited i tried to update every day. Then I lost interest (I didn't had the time too) and the patch became obsolete.
23:39:53MoosTiMiD: is it understandable, who love it :-(
23:40:10TiMiDit's not an interresting work
23:40:21TiMiDwork of a machine :)
23:40:47Moosyou don't have luck both you and Xavier
23:40:49TiMiDthat's why I didn't made a huge patch like firefly or xavier's wone
23:40:51markunTiMiD: I tried your patch again. No conflicts with the unicode patch, so that's good.
23:41:02TiMiDgreat ^^
23:41:11Mooswhen Linus have time he applied patch, work... but those weeks busy
23:41:13TiMiDI suppose you work in lower areas :p
23:42:40TiMiDI was goingto write to an indian company ;)
23:43:04TiMiDI think I'm done with mails for this evening
23:43:26Moos(reposes toi ;-) )
23:43:50XavierGrat least you should use copy and paste TiMiD!
23:45:01TiMiDdo you think I type each mail with love and tenderness ?
23:45:42TiMiDI have a template ready in my mailbox and I only have a few things to change
23:45:47XavierGrSaiyu sen jiko yo <- is this phrase translates to very important or high priority?
23:46:48TiMiDI don't know
23:46:54TiMiDare you sure for the saiyu ?
23:47:22XavierGrit is a very hilarious phrase that I hear on an anime series...
23:48:22XavierGrwell I typed it from a subtitle and yes it has the hearing of it, though I dont know if it is right
23:48:59XavierGris typing Jpanase in latin characters normal?
23:49:55TiMiDif it was saiku I could say it would mean something like "very important work" or smth like that (I'm far far away from a good japanese speaker)
23:50:28XavierGrthen what are you going to do in Japan?? Just kidding...
23:50:45TiMiDI wonder
23:52:44preglowXavierGr: pretty normal, it's called romaji
23:52:56preglowXavierGr: but mostly done by learners
23:53:10TiMiDI find it difficult to read ;(
23:53:13XavierGrThat reminds me of Greeklish
23:53:14markunlike pinjing in chinese
23:53:24TiMiDI'm most used to hiraganas and some kanjis
23:53:27XavierGrthough it's usage is only in chatrooms
23:53:42TiMiDXavierGr: why ?
23:53:53TiMiDyou can type japanes in chatrooms
23:54:02XavierGrwhy use it only in chatrooms or forums?
23:54:11XavierGror why use it at all?
23:54:40TiMiDit's UTF-8
23:54:49TiMiDyou must be in iso
23:55:21XavierGryes, what if I type this: ;
23:55:22preglowlooking at this asm again made me want a stiff drink
23:56:59amiconnHmpfz :/ My provider doesn't allow dir listings on the server when there is no idex file
23:57:39 Quit Febs ("CGI:IRC (EOF)")
23:57:42XavierGrTiMiD: The main usage of Greeklish (greek language written with latin characters) can be encountered only in chatrooms or informal Forum.
23:57:59muesli-btw we have denglish ;)
23:58:16Bagderand we have svengelska
23:58:30TiMiDand I love my french :)
23:58:32markunFriends in Colombia you Espanglish
23:58:49TiMiD(no franglish doens't exists yet :p )
23:58:52XavierGrThe main reason to use this is write a little bit faster because with only latin characters you represnt the word as it is read instead of writing it.

Previous day | Next day