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 2005-10-28

00:00:30amiconnThis is the verbatim version of the new code:
00:01:13amiconnThis is what I made from it:">
00:01:27amiconnCompare both to firmware/common/random.c ...
00:01:41Bagderstill, the license says it must be included
00:02:04Bagderbut as said, we should include the gpl license too
00:02:20TiMiDLinusN: at first I put gui_ because list_ alone can mean other things
00:02:42amiconnglist_ :)
00:02:51TiMiDexactly ^^
00:02:52*amiconn hides form LinusN
00:02:53minerLinusN: I'm using binutils-2.16
00:07:12TiMiDLinusN: so yes I intend to continue the patch until I get full support (I use only the remote ...)
00:07:28minerLinusN: Sorry about the late reply to your binutils version question, my wife came back and I had to enter reality for a while :-)
00:07:31TiMiDif you send me your patch, I can change the names as well
00:08:31amiconnBagder: We already have the gpl in docs/COPYING, but we don't include it in the .zip yet
00:11:23 Join arthax [0] (
00:12:01arthaxhello all
00:15:19arthaxThe chanel is on english only?
00:15:53arthaxok :-[
00:16:19*preglow finds out first-hand how important precision is in filter coef calculations...
00:17:22preglowseems like we'll need a Very precise sin() and cos() :/
00:20:55arthaxI would to know where are the developement of the bootloader for the H300? (question may be incorect?)
00:25:18Moosthe boot loader is still in development, and there is one guy working on it, and he is busy man
00:25:42Moosanyone know
00:26:22 Join Sandking [0] (
00:28:58arthaxthanks ;)
00:29:16Moosnp :)
00:32:05 Quit Bagder (Read error: 104 (Connection reset by peer))
00:32:28amiconnpreglow: The libmad layer 2 performance does indeed leave something to be desired
00:33:06amiconnI tested with a real-world recording now (192 kbps 48 kHz - DAB recording). 35% boost ratio...
00:33:20 Join Bagder_ [0] (
00:33:20 Join Bagder [0] (
00:33:23 Quit Bagder (Client Quit)
00:33:28 Nick Bagder_ is now known as Bagder (
00:36:33preglowamiconn: i have no idea why it is that slow
00:36:54preglowyou've got the resampler on top of that one again, of course
00:37:18preglowbut indeed
00:37:23preglowlayer2 is heaps simpler than layer3
00:37:29 Quit Sandking_ (Read error: 110 (Connection timed out))
00:37:32preglowyet it's almost just as slow as layer 3
00:37:58amiconnlayer 3 gives 6% boost here with 128 kbps, and 9% boost with 192 kbps (at 44.1 kHz)
00:39:33*preglow pines for better ways to measure performance
00:39:58preglowamiconn: there might be some cache issues here
00:40:16arthaxbye everyone ;)
00:40:19preglowamiconn: i'll see about testing out my critical function grouping one day
00:40:20 Part arthax
00:41:49amiconnI think we'll need two test cases: (1) decode and throw away data, measuring the time needed to decode the file. (2) decode to a pcm file, for testing correctness
00:42:38minermaybe i'll try getting some help again another night...bye!
00:42:47 Quit miner ("using sirc version 2.211+KSIRC/1.3.11")
00:42:50*preglow pats Slasheri on the back and whispers "waaaavwriter"
00:43:02preglowamiconn: indeed
00:43:22preglowamiconn: slasheri said it should be easy to add such functionality to rockbox now, so we don't have to hack codecs to pieces to measure their performance
00:43:30preglowbut i can't make head nor tail of playback.c
00:43:56amiconnIt's somewhat like with my mem* testing. I have a test plugin that can test correctness and measure performance
00:43:57LinusNplayback.c is not my favourite part of the rockbox source
00:44:59linuxstb_One quite easy accuracy check (at least for FLAC and lossless wavpack) would be to calculate an md5sum on the decoded data and compare that against the md5sum in the header.
00:46:08preglowlinuxstb_: yes, but that method is quite unique to the lossless codecs
00:46:24*amiconn wonders why memset on archos is slower than last time he measured
00:47:11preglowamiconn: sure, but here we've got a ton of codecs, all requiring a separate test plugin, plus we might even want to test the output of the dsp layer one day. bottom line is we need wav writing
00:47:34preglowthis wav writer should be able to be configured to do both decoding and writing, AND just decoding and tossing away the output
00:48:10preglowi know i'll sure as hell like measuring the output of my eq
00:48:25amiconnYes of course. What I want to say is that we'll probably need two mechanisms. Writing the decoded output will influence decoding speed
00:48:59preglowbut that's not the hard part
00:50:09preglowbut yeah, i'd love being able to switch on/off 1. codec operation 2. dsp operation 3. actual writing
00:50:19preglowso i can measure both codec and dsp separately
00:52:16preglowplayback.c is a bit too convoluted for me to try this
00:53:02 Quit paugh ("Leaving")
00:53:45preglowbtw, does anyone know whether the iriver fw uses 16 or 20 bit output?
00:54:03preglowi think i remember someone saying it doesn't use dma for audio, which certainly points to the latter
00:56:55LinusNpreglow: we have no choice but to use 16 bits
00:57:54*amiconn found the reason for the performance difference of memset()
00:57:57preglowLinusN: no?
00:58:33amiconnI was measuring with rombox by accident
01:00:13preglowLinusN: why's that?
01:00:43LinusNpreglow: ignore me
01:01:21LinusNyou are correct, we can use 20 bits if we use the cpu instead of dma
01:01:29preglowLinusN: you're making it hard, coming with claims like that :)
01:02:04amiconnI think that would severely hurt performance
01:03:14preglowbut i'm still wondering what iriver are doing
01:03:27amiconnHow large is an audio DMA block now? How many samples could be transferred at once w/o dma? Only one? That'd give a nice interrupt frequency...
01:03:28preglowthere are still people that claim rockbox sounds inferior
01:03:49preglowthough i'm yet to see some blind test results verifying it...
01:08:02LinusNamiconn: the fifo is 8 samples long
01:11:00preglowhard to implement?
01:11:15preglowmight be a worthwhile setting for the golden eared people
01:11:37LinusNnot too hard. methinks
01:12:19preglowat least the dsp part of it should be easy after my propsed codec sample accuracy change
01:12:45preglowproposed, even
01:12:58LinusNcould be an interesting experiment
01:13:03 Quit dpassen1 ()
01:14:28preglowi want all codecs to start outputing 28 bit data, no matter what their internal format is
01:14:35preglowso it'll be easier for dsp stuff to work on it
01:14:44preglowit'll also make arbitrary output resolution easier
01:16:16***Saving seen data "./dancer.seen"
01:30:41 Quit _aLF ("^^")
01:33:53amiconngcc really got better between my last measurements and now
01:50:01 Quit DangerousDan ("Miranda IM! Smaller, Faster, Easier.")
02:00:39LinusNook, i have now committed TiMiD's multi-screen patch
02:01:07LinusNit still needs some work, and it's possible that it will break something
02:19:10DEBUGEOF from server (Connection timed out) (snapshot: netstuff.c line 545)
02:19:10***Saving seen data "./dancer.seen"
02:19:10***Started Dancer V4.16
02:19:10***Connected to on port 6667
02:19:10***Logfile for #rockbox started
02:19:11***Server message 501: 'logbot :Unknown MODE flag'
02:19:11Mode"logbot :+i" by logbot
02:19:11 Join logbot [0] (
02:19:11 Join crwl [0] (
02:19:11 Join alxcm [0] (
02:19:11 Join Moos [0] (
02:19:11 Join Bagder [0] (
02:19:11 Join Sandking [0] (
02:19:11 Join LinusN [0] (
02:19:11 Join XavierGr [0] (
02:19:11 Join linuxstb_ [0] (
02:19:11 Join ghode|afk [0] (
02:19:11 Join Lynx_awy [0] (
02:19:11 Join lostlogic [0] (
02:19:11 Join novimon_ [0] (
02:19:11 Join guillaumh [0] (
02:19:11 Join solexx_ [0] (
02:19:11 Join amiconn [0] (
02:19:11 Join Seed [0] (
02:19:11 Join ]Maggie[ [0] (
02:19:11 Join Zagor [0] (n=bjst@pdpc/supporter/sustaining/Zagor)
02:19:11 Join ryan_j [0] (
02:19:11 Join markun [0] (
02:19:11 Join pabs [0] (
02:19:11 Join Rick [0] (i=rick@unaffiliated/Rick)
02:19:11 Join CoCoLUS [0] (
02:19:11 Join ze [0] (
02:19:11 Join goa [0] (
02:19:11 Join Slasheri [0] (
02:19:11 Join Ismo [0] (
02:19:11 Join phaedrus961 [0] (
02:19:11 Join merbanan [0] (
02:19:11 Join Strath [0] (
02:19:11 Join pill [0] (
02:19:11 Join HCl [0] (
02:19:11 Join yosemite [0] (
02:19:11 Join crash_ [0] (
02:19:11 Join Nilisco [0] (
02:19:11 Join preglow [0] (
02:19:11 Join Hadaka [0] (
02:19:11 Join TiMiD [0] (n=TiMiD[
02:19:11 Join mbr [0] (
02:19:11CtcpVersion from freenode-connect!freenode@freenode/bot/connect
02:19:13LinusNno large commit without an "oops"
02:20:02linuxstb_LinusN: That compiles fine now for the H120/H140.
02:20:49LinusNi haven't dared to try with logf support...
02:21:16*linuxstb_ dares...
02:26:34TiMiDstill "some" compile errors :)
02:27:13LinusNTiMiD: that's from the alac commit
02:27:22LinusNmy commit is building now
02:33:51MoosTime to go to sleep here, have a good night all (thanks Linus for the patch)
02:34:33 Quit Moos ("Glory to Rockbox")
02:36:03TiMiDseems that it's compiling cleanly :)
02:38:32LinusNok, the ondio has problems
02:41:52markunAnd no conflicts with the unicode patch still
02:42:12TiMiDhmm strange
02:42:28markunthe viewer is a bit buggy with unicode
02:42:31TiMiDI just compiled a cvs snapshot and it works fine with ondio on sim
02:42:46LinusNTiMiD: sim is not the same as target
02:43:10TiMiDso I realy need a sh1 gcc compiler
02:44:41TiMiD(but this will be for tomorrow :)
02:53:29LinusNfixes committed
02:55:40TiMiDI saw :)
02:56:42TiMiDis the warning on fm recorder fixed too ?
02:58:53TiMiDalso should I change all the ++i to i++ and remove the == true things
02:59:11LinusNi removed all ==false and ==true i could find
02:59:46LinusNi would appreciate changing the preincrements, but it isn't exactly top-prio
03:00:31TiMiDI can do it tomorrow when changing some fn names which are too long
03:00:45TiMiDthen I could start to implement menus and so on :)
03:01:09TiMiDI hope I won't break anything with CVS
03:01:20LinusNTiMiD: don't worry
03:02:26TiMiDI want to try :)
03:02:51TiMiDmaybe should I try to commit the starfield sim before I go to bed ?
03:03:19TiMiD(even if it's very useless)
03:05:55TiMiDpoor server it has been compiling a lot this evening ^^
03:07:28LinusNyes... :-)
03:07:57TiMiDI try it on other targets then I try to commit (Hope this won't make the server explode ^^)
03:08:49LinusNwee, green build
03:09:10TiMiDhuhuhu :)
03:22:41XavierGrhmm maybe time to update my FM patch. Linus will you have time to check it tomorrow not that the remote stuff is up to TiMiD?
03:22:49XavierGrnot = now
03:23:11LinusNi hope so
03:23:43XavierGrI made a neat Icon too, it is like the configuration I con but with a R inscribed in it.
03:26:40 Join Sandking_ [0] (
03:29:27XavierGris it okay when - I apply a patch - that sometimes it says: Hunk #1 succeeded at 94 (offset 6 lines).
03:29:38XavierGrIs this reliable or maybe there is an error?
03:29:54XavierGrBecause if the patch is clean and good it never says anything like that to me.
03:30:03LinusNit's mostly ok
03:30:30LinusNdoesn't hurt to check, though
03:30:45XavierGrby compiling you mean, right?
03:31:07LinusNyes, and checking in a text editor
03:35:09LinusNtime to call it a day
03:35:12LinusNnite all!
03:35:19 Part LinusN
03:36:30TiMiDstarfield commited :p
03:37:10amiconnHope this doesn't break usb power status display etc :/
03:38:27TiMiDI don't think a plugin can break anything ;)
03:39:18amiconnNo, I mean the multi lcd support
03:39:25TiMiDI know :D
03:39:43TiMiDI can't test since I don't have the unit in my hands ....
03:40:01amiconnLinus last commit looks somewhat suspicious to me :/
03:40:25XavierGrSukoi!! My patch still applies cleanly. Now I will upload the latest version to the tracker.
03:40:33amiconnMaybe this is false alarm...
03:40:35TiMiDthat " Fixed Ondio and V2/FM compiling problems from the multiscreen patch" ?
03:40:50 Quit Sandking (Read error: 110 (Connection timed out))
03:42:14TiMiDwhat seems suspect ?
03:43:56amiconnI need to test this stuff on all my boxes...
03:44:27amiconnOoops, but not right now. It got late... :/
03:44:34TiMiDI think so :)
03:44:51TiMiDI'm going to bed right now !
03:45:04*amiconn was forcing himself to work on memcpy
03:45:15TiMiDyou should sleep :)
03:45:35TiMiD(when I read the logs I wonder when do you sleep ;) )
03:47:20*amiconn managed to make it a bit slower for very small sizes :(
03:48:09TiMiDgood night !
03:48:30XavierGrbye amiconn!
03:57:35XavierGrOkay, patch updated! Linus if you find some time tomorrow please have a look and inform me for any further points that you want me to fix.
04:00:49 Join webguest05 [0] (
04:01:37XavierGrTiMiD: Gongratulations too, (for the CVS access) lucky you ;P
04:01:56 Quit webguest05 (Client Quit)
04:03:18 Join Midgey34 [0] (
04:19:14***Saving seen data "./dancer.seen"
04:50:27 Quit Midgey34 ("Chatzilla [Firefox 1.0.7/20051010]")
06:19:16***Saving seen data "./dancer.seen"
06:25:56 Join Sandking [0] (
06:40:31 Quit Sandking_ (Read error: 110 (Connection timed out))
07:02:28 Join amiconn_ [0] (
07:17:06 Quit Sandking ()
07:20:57 Quit amiconn (Read error: 110 (Connection timed out))
07:20:57 Nick amiconn_ is now known as amiconn (
07:45:39 Join _FireFly_ [0] (
07:46:24_FireFly_moin :)
07:48:21_FireFly_ok i have an updated version for my patch against the latest cvs :)
07:57:34_FireFly_it can be found here
07:59:32 Quit _FireFly_ ("Leaving")
08:00:45 Join B4gder [0] (n=daniel@
08:03:55 Join mbr_ [0] (
08:15:04 Join ender` [0] (i=ychat@
08:19:19***Saving seen data "./dancer.seen"
08:19:46 Quit mbr (Read error: 110 (Connection timed out))
08:22:25 Quit lostlogic (Remote closed the connection)
08:23:06 Join lostlogic [0] (
08:43:19 Join _FireFly_ [0] (
08:49:03_FireFly_ (bugfix for entering the quick-menu from the filetree when long pressing the remote-mode-button )
08:59:19 Nick Lynx_awy is now known as Lynx_ (
09:07:43 Quit solexx_ (Read error: 104 (Connection reset by peer))
09:09:31 Join einhirn [0] (
09:14:55 Join cYmen [0] (
09:20:06 Join cYmen_ [0] (
09:23:34 Join solexx [0] (
09:25:16 Join cYmen__ [0] (
09:26:59 Quit _FireFly_ ("IceChat - Its what Cool People use")
09:27:58 Join _FireFly_ [0] (
09:30:31 Join cYmen___ [0] (
09:30:31***Alert Mode level 1
09:30:31DBUGEnqueued KICK cYmen
09:30:31DBUGEnqueued KICK cYmen_
09:30:31***Alert Mode level 2
09:30:31DBUGEnqueued KICK cYmen__
09:30:31DBUGEnqueued KICK cYmen___
09:30:31***Alert Mode level 3
09:32:38 Quit cYmen (Connection timed out)
09:35:40 Join cYmen [0] (
09:35:40***Alert Mode level 4
09:35:40***Alert Mode level 5
09:35:40DBUGEnqueued KICK cYmen
09:35:40***Alert Mode level 6
09:35:40***Alert Mode level 7
09:35:40***Alert Mode level 8
09:37:47 Quit cYmen_ (Connection timed out)
09:42:56 Quit cYmen__ (Connection timed out)
09:45:41***Alert Mode OFF
09:48:01 Quit cYmen___ (Connection timed out)
10:19:20***Saving seen data "./dancer.seen"
10:47:06 Join dwihno [0] (n=dw@
10:55:52 Nick mbr_ is now known as mbr (
11:00:03 Join ashridah [0] (
11:40:46 Join Edunn [0] (
11:41:26 Part Edunn
11:49:38 Join DangerousDan [0] (
12:01:30 Join ripnetuk [0] (
12:01:36ripnetukyay multi screens :) thanks guys
12:03:22 Quit ripnetuk (Client Quit)
12:09:29 Join RiverFish [0] (
12:09:54 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
12:12:51 Quit RiverFish (Client Quit)
12:12:54 Join RiverFish [0] (
12:16:04RiverFishI think I read someone saying that all the patch integration is being done by one person
12:16:14RiverFishIs there an opportunity to help out there?
12:16:25 Join ripnetuk [0] (
12:16:29RiverFishSeems like a large load for one person
12:16:40preglownot true
12:16:53RiverFishok i stand corrected
12:16:55preglowpatches are integrated by anyone with enough skill to properly verify they're ok
12:19:23***Saving seen data "./dancer.seen"
12:23:14linuxstb_Morning all. Minor point, but should source files be iso-8859-1, UTF-8, or just plain ASCII ? Or is it not an issue?
12:24:14preglowascii, i guess
12:24:26preglowdo you need any non-ascii symbols?
12:24:36B4gderthey are 8859-1 atm
12:25:01B4gderZagor's first name being the letter that shows it ;-)
12:25:06B4gderusing even
12:41:04linuxstb_I don't need them - it's just that TiMiD's files have his name with an accent - sometimes 8859-1, sometimes utf-8
12:41:22linuxstb_The GUI files are 8859-1, starfield.c is utf-8.
12:41:45B4gderthat's TiMiD's commit, right?
12:41:58linuxstb_As I said, not a big deal, just asking.
12:42:01preglowlinuxstb_: any closer to understanding the dbpoweramp thing?
12:42:22linuxstb_preglow: No - I was playing with ALAC last night.
12:42:55linuxstb_But I think there is more general cleaning that needs doing with libffmpegFLAC - that may find the problem. i.e. making sure that variables are not used unintialised.
12:44:10preglowit's small enough to be managable, at least
12:45:20linuxstb_I think the fact that the memset had a big effect implies an uninitialised variable problem.
12:45:57 Join Moos [0] (
12:47:22linuxstb_Did you see the ALAC commit? It's now outputting 28-bit samples ready for your DSP. I also managed to remove two 16KB buffers - but that didn't give much of a speed boost.
12:47:54preglowyeah, i saw
12:47:59preglownice still
12:48:48preglowjust hope i'll get this bloody eq working soon
12:49:25linuxstb_I'm not sure about changing the other codecs. I can probably handle WAV and maybe wavpack, but I don't want to touch the lossy codecs.
12:49:51preglowno worries, i'll probably deal with it myself
12:50:16linuxstb_Also, how should we be handling mono? At the moment, a lot of the codecs (all?) convert to stereo before sending to the DSP - obviously a bad thing.
12:50:25preglowhell yes
12:50:31preglowmusepack does that, i know
12:50:34preglowalso alac
12:50:46preglowand wavpack, in the codec plugin
12:51:05linuxstb_Yes - I think I've done that with all the codecs I've implemented. I want to fix it.
12:51:36linuxstb_Does the DSP handle a mono stream at the moment?
12:53:33preglowit should
12:55:20linuxstb_Looking at dsp.c, it looks like I should pass the same array pointer twice to pcmbuf_insert_split()
12:55:31linuxstb_Is that the right thing?
12:57:40preglowi think you can pass whatever you want, it'll get ignored anyway
12:57:45preglowbut same twice is probably ok
12:58:14preglowit would be nicer if it took a long** as parameter
12:58:25preglowbut doesn't matter that much
12:59:12linuxstb_If the second pointer isn't used, then I should probably pass NULL.
12:59:22linuxstb_Forget that.
12:59:43linuxstb_I'll just pass the normal left and right arrays.
13:00:00linuxstb_I'll give it a go later and see what happens.
13:12:59ripnetukdoes the latest iRiver boot still always load the original iRiver firmware when switched on from the remote?
13:14:07ripnetukso if i flash the latest bootloader, it will always boot rockbox when I turn it on from remote? atm it boots iRiver if turned on from remote, and rockbox if turned on from unit (unless rec presses)
13:15:39ashridahboots rockbox from both atm
13:15:48ashridahthere was a period when it didn;'t do that.
13:15:51linuxstb_ripnetuk: I don't have a remote, but I'm sure the bootloader will do what you want. That's the logical thing to happen. I'm guessing it was a bug in earlier bootloaders.
13:16:21ripnetukcool thanks :) now flashing :)
13:16:43ripnetuki dont think it was a bug - i think it was a deliberate feature before rockbox on iRiver was properly usable...
13:16:49ripnetuktoday is the day i switch over for good
13:17:12linuxstb_Why so long? I'm guessing you rely on the remote?
13:17:30ripnetukcompeltely... i should have got a iAudio :)
13:17:46preglowlinuxstb_: no, not a bug, it was requested
13:18:05ripnetukdamn it boots quick now :)
13:18:05linuxstb_preglow: OK :). But then it was unrequested?
13:18:20ripnetuki can confirm it behaves correctly (ie, boots rockbox)
13:18:28 Join DrMoos [0] (
13:18:36 Quit Moos (Read error: 104 (Connection reset by peer))
13:18:43 Nick DrMoos is now known as Moos (
13:18:56preglowlinuxstb_: we reverted the feature when the remote actually started working in rockbox
13:19:24linuxstb_OK. Must have all happened during my period of absence.
13:19:30preglowit did
13:19:42preglowman, i wonder how i'm going to make this eq work without killing the precision totally
13:20:16preglowi should have started fiddling with fixed point filters earlier
13:20:41linuxstb_Is it all written in asm?
13:20:46preglowno, all c at the moment
13:20:57preglowthe only thing i'm going to write in asm is the filtering loop itself
13:20:59preglowand that's not big
13:21:00linuxstb_That's good.
13:21:20preglowthe coef calculation is in pure c, using 64 bit arithmetic
13:24:26 Join Quelsaruk [0] (
13:24:38preglowlook who's back
13:24:45Quelsaruk(after serveral months)
13:24:49Quelsarukwork is too hard
13:25:10B4gderbuenas dias Quelsaruk
13:25:19Quelsarukhej B4gder
13:25:37Quelsarukhmm... mur da bra? (or similar... my swedish is quite poor as you know)
13:25:54B4gderhehe, I'm fine but overloaded with work
13:26:05Quelsaruklike me :(
13:26:32Quelsaruki have installed bleeding edge version, and get a strange error
13:26:45Quelsaruk"filetype array full"
13:27:09Quelsaruk(oh, before that i have flashed bootbox)
13:28:39ripnetukwow :) bookmarks on my iRiver
13:29:13QuelsarukB4gder: should i kill a goat? or is there any functional way to fix that?
13:29:28linuxstb_Are you using the standard viewers.config ? Do you have lots of obsolete plugins in your plugins directory?
13:29:43B4gderQuelsaruk: I'd suggest you wipe your .rockbox first and then re-install
13:29:50linuxstb_About to say the same thing.
13:29:53Quelsarukok, deleting .rockbox then
13:30:27Quelsaruki tend to overwrite...
13:30:32Quelsaruki should delete
13:37:59ripnetukare on the fly playlists (ie, right'click on a dir, insert into playlist) persistent?
13:38:06ripnetukie, will they survive a reboot?
13:39:40CtcpIgnored 5 channel CTCP requests in 5 minutes and 9 seconds at the last flood
13:39:40*ripnetuk rtfm
13:39:54Moosresume do you mean?
13:40:05Quelsarukbleeding edge builds doesn't have a rombox.ucl file?
13:43:25ashridahhrm, can't find any bugs about it
13:43:29ashridaharghl. wrong chnanel
13:46:12Quelsaruki'll be back... someday :d
13:46:24 Quit Quelsaruk ("KVIrc 'Marmalade'")
14:01:03 Join muesli- [0] (
14:14:15 Join amiconn_ [0] (
14:19:26***Saving seen data "./dancer.seen"
14:21:16muesli-re (c)
14:21:26markunHey muesli-
14:21:46muesli-hossa markun
14:22:02 Quit amiconn (Read error: 110 (Connection timed out))
14:22:03 Nick amiconn_ is now known as amiconn (
14:22:31markunamiconn: with TiMiD's patch applied the unicode patch makes rockbox too big for the player v1 :(
14:23:28markunI've been looking at the linux kernel source a bit and googling some to get an idea of compressing rockbox with gzip
14:23:57linuxstb_I thought TiMiD's patch would mean an eventual reduction in code size - when it's fully implemented and dead code is removed.
14:24:14linuxstb_I'm not sure, but I think the current patch is the peak in terms of code size, and further patches will reduce it.
14:25:16 Quit Zagor ("Client exiting")
14:25:40markun21.57.59 # <TiMiD> we will gain some size with menu and playlist
14:26:21preglowmarkun: why compress with gzip? doesn't rockbox for archos already have code to deal with ucl compression?
14:26:32markunWill we gain more space or will it get bigger in size?
14:26:41markunpreglow: Didn't know that.
14:27:39markunpreglow: So is rockbox already compressed on the player v1?
14:27:45preglowit's compressed in rombox
14:27:59B4gderthe size problem is really for the non-flashed units
14:28:54SlasheriHmm, what causes that size limit?
14:29:27B4gderthe archos firmware
14:29:41B4gdersince it loads and runs the image from disk
14:29:51B4gderand it refuses to do so if its too big
14:30:07ripnetuki think it would be good if you could browse playlists in a proper treeview like thing, ie, as if the filesystem contained only the files in the list. That way, we could have one playlist containing everything, and manually choose a tune by browsing the tree, but keeping the playlist active, so we can go back to shuffle everything easily.
14:30:08Slasherioh, but that's not so bad
14:30:24Slasheribecause you still can flash rockbox or bootloader?
14:30:29ripnetukb4gder - cant we just have a second stage loader?
14:30:32B4gderSlasheri: not all players are flashable
14:30:38B4gderripnetuk: yes we can
14:30:43ripnetukjust a mini rolo
14:31:26Slasheribzip2 compression could be quite effective also
14:31:36Slasheribut maybe slow
14:31:53ripnetukor even better, if you choose a track manually that is in the playlist, it should just to that point in playlist, instead of creating a dynamic playlist for that dir. not sure how fast that could be done
14:32:31ripnetukkinda like follow playlist in reverse
14:32:32markunOr, but it's similar to bzip2
14:38:56 Join einhirn [0] (
14:41:15preglowSlasheri: bzip2 decompression is more than slow enough to probably have an impact on our boot time, heh
14:42:29Slasheripreglow: Hmm.. Then we could implement a 2-stage boot-loader..
14:42:53SlasheriThe first part would be very minimal and it would access the disk and load the image
14:43:32preglowwe could NOT use bzip2 :P
14:43:40ripnetukor mod the original archos fw to remove the limit
14:43:41preglowi like ucl
14:43:46Slasherihehe, true :D
14:43:51preglowfast and small decompressor
14:44:03preglowripnetuk: some units aren't flashable
14:44:13ripnetuki see, no go then
14:44:30preglowSlasheri: you got a wav writer coming up some time soon? :>
14:44:37ripnetukis it ok to post feature requests to users mailing list, without some example code / patch?
14:45:03preglowi tried having a go at one about a week ago, but the playback code was too intimidating
14:45:09Slasheripreglow: hehe, currently i am quite busy the next month (designing and programming a pcb with a strict dead line) :/
14:45:26Slasheripreglow: but maybe some day, when i have time, i could write it :)
14:45:35preglowwould be truly nice
14:48:45 Join webguest78 [0] (
14:50:46webguest78huy guys, just want to say, ROCKBOX ROCKS! i just wish i could help more by developing, but alas, im a noob! question: how is rockbox video implementation going?
14:51:08muesli-preglow ;) your part ;)
14:51:43B4gderwebguest78: there is no work being done on that, afaik
14:51:56B4gderif you mean on iriver
14:52:12webguest78ah right, but it is possible? (yes im talking about iriver)
14:52:15ripnetukthe display is very slow on iRiver, not sure how good it will be for vid
14:52:28ripnetukslow as in slow refresh rate, so trails would exist
14:53:21preglowmuesli-: hah, you needn't be afraid of me coding video codecs any time soon
14:53:22webguest78is the irivers' display slower than the archos' and ipods?
14:53:34B4gderslowe than the archos ones for sure
14:53:41ripnetuknot sure about ipos, but as far as i can remember slower than archos
14:53:49ripnetukbut never owned both units at same time
14:54:05B4gderthere's a significant difference
14:54:19webguest78i mention the ipod because of ipod linux
14:54:29webguest78and its video capability
14:54:56linuxstb_The ipodlinux video capability is simply playing back uncompressed video and uncompressed audio. The files are huge.
14:55:07B4gderbut does the movies look good?
14:55:11linuxstb_I have no idea.
14:55:41linuxstb_ipodlinux is now (almost) working on my ipod. So I'll try it out soon. Problem is that each major app has its own LCD code that needs updating.
14:55:46webguest78well from the sample video it didnt look too bad, obviosuly a gimmick rather than video with a practicle purpose.
14:56:14preglowlinuxstb_: each _APP_ ?
14:56:15B4gderthere is video with practible purposes?
14:56:18webguest78i was amazed what you guys did with the JPEG veiwer, amazing really.
14:56:37preglowlinuxstb_: the driver isn't in the kernel?
14:56:42webguest78practicle, as in, watching a film or whatever lol
14:57:19linuxstb_There is a character-based framebuffer driver in the kernel. For graphics, I think podzilla uses microwindows (hence a need for a microwindows driver). Other apps may just use the LCD directly with no driver.
14:57:37linuxstb_This is one problem with IPL - the development is very fragmented.
14:58:16linuxstb_It is very much an operating system and various independent applications. Unlike Rockbox's "all in one" approach.
14:59:37webguest78does the midi2wav application in the iriver actually work? i have to reset.
14:59:58 Join Nico_FR [0] (
15:00:08 Quit Nico_FR (Client Quit)
15:00:44 Join Nico_FR [0] (
15:00:55webguest78are you guys in need of any hardware at the moment? as in iriver parts etc?
15:01:18B4gderwe're mostly in need of people with time and skills
15:01:22linuxstb_I could do with an 80GB hard drive :)
15:01:38webguest78i have a mainboard minus the LCD if anyone wants it (h140)
15:02:23linuxstb_I think there's a Wiki page for buying/selling/giving away parts.
15:02:23webguest78im not sure if the LCD can be replaced though.
15:04:44webguest78well ill kep doing the bleeding edges and searching for bugs:) good luck with the project, great work.
15:05:47 Quit Nico_FR ()
15:13:09 Quit webguest78 ("CGI:IRC (EOF)")
15:13:18 Join _DangerousDan [0] (
15:17:06 Quit DangerousDan (Read error: 110 (Connection timed out))
15:17:17 Join XavierGr_ [0] (
15:37:23 Quit ripnetuk ("Ninja IRC v1.5.8.1(#1) exiting after 3h20m59s of use")
15:39:14 Quit XavierGr (Read error: 110 (Connection timed out))
15:48:21 Join XavierGr [0] (
15:51:53 Join ripnetuk [0] (
15:52:07 Join Nico_FR [0] (
15:53:01 Quit Nico_FR (Client Quit)
15:53:37 Join webguest84 [0] (
15:54:11webguest84is the ipod's hardware (CPU, RAM etc) more or less powerful than the irivers'?
15:54:50B4gderram is the same, on the cpu side I guess the ipod is somewhat more powerful
15:55:07B4gderguess because its a different dual-core cpu
15:55:19webguest84ah, thanks
15:56:06 Quit B4gder ("time to say moo")
15:56:54 Quit ripnetuk ("Ninja IRC v1.5.8.1(#1) exiting after 5m4s of use")
15:56:59 Join ripnetuk [0] (
15:57:34ripnetukTiMiD - u there?
15:57:48ripnetuki still havent managed to break your remote patch
16:02:26 Quit _DangerousDan ("Miranda IM! Smaller, Faster, Easier.")
16:05:05 Join DangerousDan [0] (
16:05:38linuxstb_webguest84: The iPod has two 75MHz ARM processors. The irivers have a single 124MHz Coldfire. They both have 96KB of fast internal RAM and 32MB of slower SDRAM.
16:06:14linuxstb_The new iPod video also has a very powerful video processor with its own pair of 75MHz (I think) processors.
16:13:39 Quit XavierGr_ (Read error: 110 (Connection timed out))
16:13:50webguest84thanks :)
16:14:42 Quit webguest84 ("CGI:IRC")
16:19:28***Saving seen data "./dancer.seen"
16:19:44 Join hshah [0] (
16:19:51hshahheya all
16:19:57hshahwhat does the multicreen thing do?
16:20:44ashridahremote support for iriver, my guess.
16:21:13hshahoh really....
16:21:25ashridahdoesn't support everything yet, WPS doesn't work on the remote, but the file tree does, and it's the right approach, so it's not going to be a mess to implement
16:21:34ashridahor so i gather
16:21:42ripnetukit works good
16:21:48ripnetukthe remote support i mean
16:21:49hshahwell yay
16:23:44ripnetukhas anyone successfully built the sim on FC4?
16:24:46TiMiDripnetuk:I'm here (for a short moment)
16:24:53TiMiDwhat about my patch ?
16:25:57ripnetuki love it :) thank you very much
16:26:12ripnetukbeen using it all day, and been 100% stable
16:27:54linuxstb_ripnetuk: The sim just needs the basic xlibs - so it should work everywhere. Are you having problems?
16:28:16preglowlinuxstb_: what is bascially left before you've got a working ipod port?
16:28:26preglowlinuxstb_: have you dealth with the lower level stuff of interrupt handlers and such?
16:28:35 Quit hshah ("CGI:IRC")
16:28:56markunTiMiD: I have a question about what you said last night. Will the size increase or decrease with WPS and menu?
16:29:03linuxstb_No - interrupt handlers are probably the next thing to do (before I finish ATA). I then need to finish ATA and write a proper button driver.
16:29:23linuxstb_It's just that I've not really spent much time on it in the last few weeks.
16:30:18TiMiDmarkun: I can't tell
16:30:43TiMiDbut it's sure that the size will decrease with menu and playlist implementation
16:31:07markunSo you CAN tell :)
16:31:46ripnetuki think the sim has a symobl that clashes with something in fc4
16:31:52ripnetuki get /usr/bin/ld: errno: TLS definition in /lib/ section .tbss mismatches non-TLS reference in /projects/rockbox/cvs/rockbox/build/sim/librockbox.a(errno.o)
16:31:52ripnetukcollect2: ld returned 1 exit status
16:32:10linuxstb_preglow: I'm working on a bootloader, which is similar to the h120/h140 bootloader. i.e. it is a very stripped down version of Rockbox using the same ATA driver, LCD driver etc. Once the bootloader is working (i.e. loading the image from disk), then a basic Rockbox is almost done as well.
16:32:21TiMiDripnetuk: you have the same problem that I had some time ago
16:32:29preglowlinuxstb_: i'm most certainly looking forward to it
16:32:37ripnetuktimid - how did u get around it
16:32:44TiMiDsince I run gentoo it was not very difficult to solve
16:33:05TiMiDI just had to recompile glibc with both linuxthreads and pthread support
16:33:09linuxstb_From looking at the recent IPL patches, the Nano is almost identical to my 60GB iPod Color - so the Rockbox changes will be minimal.
16:33:15TiMiDbut on fc4 I doubt this is possible
16:33:20ripnetukno :)
16:33:28TiMiD(it is but ...)
16:33:35ripnetuki thought it might be something to do with the fact im running Xen
16:33:42ripnetukwhich doesnt play nice with TLS
16:33:52ripnetukbut it seems others have had similar issues with FC4 and compiling PHP
16:34:42ripnetukoh well, i need to just build another virtual machine on xen - i think i will use debian... any known issues with debian unstable?
16:34:43TiMiDlook at the gentoo ebuild if you want to recompile, the correct options are given
16:35:04ripnetuki think for me it will be easier to make a new VM to play rockbox in
16:35:28ripnetukor maybe (shudder) compile for win32 :)
16:36:04TiMiDthat's the first solution I choosed : win32
16:36:19TiMiDbut developping under linux is much more convenient ;)
16:36:24ripnetukyes much more so
16:36:37ripnetukespecially if you wanna do it @ work via a vnc/nx session :)
16:37:54linuxstb_You can cross-compile the win32 sim under Linux and then run it using Wine.
16:38:18ripnetuk:) cool
16:38:30ripnetukdoes the sim display the remote?
16:38:58TiMiDyes it displays it :)
16:39:05TiMiDyou want to develop remote too ?
16:39:22TiMiDnice :p
16:39:23ripnetuki would like to help, maybe try a plugin or 2
16:39:37TiMiDoh just plugins !
16:39:41ripnetukbut, c is not my first language (delphi / c# is)
16:39:51TiMiD(btw, I started there : developping plugins)
16:39:55ripnetuki was going to start with a plugin.
16:40:08ripnetukits only menus and wps thats missing atm isnt it?
16:40:14TiMiDone thing that would be very nice to do with plugins
16:40:20ripnetukwhats that?
16:40:34TiMiDwould be to add the stop button on the remote to quit
16:41:01ripnetukyes. im going on holiday for a week tomorow, so wont be able to play until my return.
16:41:09TiMiDbecause when you loanche a viewer or a plugin from remote you are obliged to take your player out of your bag which is not very convenient ;)
16:41:10ripnetukbut i want to get the sim compiling...
16:41:51ripnetukalthough i dont really use plugins much
16:42:54*ripnetuk is gonna go and show /usr/bin/ld who is boss :)
16:43:50TiMiDfor example
16:44:04TiMiDI have jpeg files in my album folder
16:44:09TiMiDand txt for the lyrics
16:44:47TiMiDwhen I'm using remote, if I mistake and select a file that is not audio
16:44:50TiMiDI'm dead ;)
16:45:41ripnetukhehe i bet you are adept at getting out of it 'blind' tho ?
16:46:02TiMiDsure :)
16:46:32TiMiDI use it mainly on my bike so I have the choice beteen falling or selecting the wrong file :)
16:47:22ripnetukhehe i use mine on a bike too - can you do that thing where u stop by a lamppost and balance (no feet down) and fiddle with your mp3 player / phone :)
16:49:45TiMiDlast year I could do it while riding, even if I was not stopped
16:50:45TiMiDbut my wheel gets twisted so I must hold it with one hand now I can't ride without hands
16:51:15 Quit ashridah ("sleep")
16:52:11 Quit ghode|afk (Read error: 110 (Connection timed out))
16:52:21ripnetukyes. its very dangerous trying to interact with gadgets while cycling
16:52:33ripnetukin fact thats the only reason i upgraded from archos to iriver - the remote
16:54:34muesli-yeah...remote roxxx
16:54:53TiMiDI have to go
16:55:05TiMiDI will be back in 1h
16:55:15TiMiDthen I will start working on remote again :)
16:55:25muesli-cheers TiMiD
16:55:34ripnetukcool... i look forward to seeing your next achivement :)
16:55:43muesli-you're the most beautiful dev ive ever known ;)
16:56:41TiMiDmuesli-: sry, but ... I prefer girls you know ... :P
16:57:01muesli-we should edit your start up routine ;)
16:58:12ripnetukdo we already have a .wps code for disk activity? it would be cool to reflect the red LED on the LCD
16:58:50muesli-anyway..gtg too
16:58:51TiMiDI don't think wps can do thi kind of updates
17:05:06 Quit ripnetuk ("Ninja IRC v1.5.8.1(#1) exiting after 1h8m10s of use")
17:09:38 Quit XavierGr (Read error: 110 (Connection timed out))
17:12:39 Join linuxstb [0] (
17:20:26 Quit muesli- (Read error: 110 (Connection timed out))
17:25:10 Quit linuxstb_ (Read error: 113 (No route to host))
17:35:00 Join linuxstb_ [0] (
17:36:22 Quit linuxstb (Read error: 110 (Connection timed out))
17:37:04amiconnTiMiD: The gui code adds quite something to the binary size, 2288..4464 bytes depending on target
17:38:53 Nick linuxstb_ is now known as linuxstb (
17:44:36 Join ripnetuk [0] (
17:45:44ripnetuki got the sim building under fc4 :)
17:50:41 Join guillauh [0] (
17:51:48 Quit ripnetuk ("Ninja IRC v1.5.8.1(#1) exiting after 7m44s of use")
17:51:49amiconnripnetuk: We already have code that can simulate a LED, not for wps but for status bar.
18:04:55 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
18:08:18 Quit guillaumh (Read error: 110 (Connection timed out))
18:08:22 Join gromit` [0] (n=gromit`
18:08:47 Nick Lynx_ is now known as Lynx_awy (
18:12:55 Join XavierGr [0] (
18:15:38 Join einhirn [0] (
18:18:28 Join dpassen1 [0] (
18:19:29***Saving seen data "./dancer.seen"
18:25:34 Quit _FireFly_ (Read error: 104 (Connection reset by peer))
18:25:44 Join paugh [0] (n=kickback@2001:5c0:8fff:ffff:8000:0:3e03:6822)
18:34:15 Join ghode|afk [0] (
18:40:37 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
19:04:33 Quit Moos (Read error: 104 (Connection reset by peer))
19:04:35 Join DrMoos [0] (
19:04:58 Nick DrMoos is now known as Moos (
19:10:19 Part linuxstb ("Leaving")
19:16:07 Quit RiverFish ("CGI:IRC (Ping timeout)")
19:18:07 Join linuxstb [0] (
19:22:42 Quit XavierGr (Read error: 110 (Connection timed out))
20:18:23amiconnBagder: r u there?
20:19:31***Saving seen data "./dancer.seen"
20:22:30 Join Kohlrabi [0] (
20:27:41TiMiDamiconn: yes
20:27:45TiMiDI'm here
20:27:55amiconnI found some bugs in the multi-display code :(
20:28:21amiconn(1) The stop-on-wrap-when repeat doesn't work any more
20:29:40amiconn(2) When returning from a plugin that sets the default font, the font isn't immediately reset to the ui font. That happens as soon as you scroll one up or down
20:29:59amiconnLooks somewhat funny
20:30:28amiconnAnother detail is that the center-scroll margin is different from what it was
20:30:39preglowand behaviour
20:30:50preglowdidn't it use to stop when you hit the ends of the list+
20:30:56preglownow it just continues on to the top or bottom
20:31:03amiconnpreglow: That is (1) in my bug list...
20:31:11TiMiDwhat do you mean by stop-on-wrap ?
20:31:14preglowwhat i said
20:31:27preglowwhen you press up or down to scroll a list
20:31:34preglowthe scrolling used to stop when you hit top or bottom
20:31:55preglowjust try any old build of rockbox and try it yourself
20:32:04TiMiDthat's not a bug
20:32:08TiMiDit's wanted
20:32:14amiconnI consider it a bugf
20:32:18preglowwell, says who?
20:32:19TiMiDI prefer it this way but if you have objections
20:32:30preglow think you should retain it so nothing's changed after your patch
20:32:37preglowand THEN you can start changing it
20:32:59amiconnTiMiD: You can test (2) with any plugin that sets the font, e.g. cube
20:32:59preglowif people want you to, of course
20:33:12TiMiDamiconn: I believe you :)
20:33:14dpassen1i think the old behavior was nice, as well
20:33:26TiMiDbut the plugin should set the font as it was before
20:33:32preglowhmm, no
20:33:44preglowthat's easier done by the widget code, if you ask me
20:33:54preglowor whatever loaded the plugin
20:34:03TiMiDit can be also fixed
20:34:05preglowcode one place is better than the same code done in a dozen places
20:34:53TiMiDthe problem is that my code will spent his time to fix the fonts
20:35:32TiMiDI can include it
20:35:36amiconnTiMiD: The code size is another problem... I hope your expectation will be come true and the size will decrease again, hopefully enough to make unicode fit as well...
20:36:02TiMiDamiconn: just consider that a lot of code is present twice right now
20:36:15TiMiDlike status bar, button bar etc
20:36:49TiMiDbut I can't remove them because a lot of code still rely on them
20:37:16preglowTiMiD: there's not much time involved there, the old rockbox did it without anyone complained
20:38:09TiMiDpreglow: about what ?
20:38:19TiMiD(I fixed the font btw)
20:39:04preglow20:34 < TiMiD> the problem is that my code will spent his time to fix the fonts
20:39:07preglowi was talking about that
20:39:23TiMiDI fixed it
20:39:41TiMiDshould work
20:39:54TiMiDsince you redraw the screen after each font change
20:40:06TiMiDI made the change in the fn that handles drawing
20:41:30linuxstbThis FLAC bug is proving hard to track down.
20:41:49TiMiDabout the scroll that stops when reaching the bottom of the list ... I don't like it
20:41:52preglowi can still do some remote debugging for you if the logd feature still works
20:41:58TiMiDmaybe I should submit a poll ?
20:42:01linuxstbThe file in question stops decoding at frame 17. But there is nothing unusual about frame 17 - it's the encoded the same as previous frames.
20:42:29linuxstbIt's OK - it doesn't crash any more, so I can use logf and check it when the codec stops.
20:42:32dpassen1seems like one of those things that can be an option, but i definitely preferred the old stop at bottom behavior
20:42:33TiMiDon the forum to know what is the most wanted
20:42:59TiMiDI like the new way because you can go straight to the botom of the list
20:43:08TiMiDif you hav lot of files in a dir
20:43:12amiconnTiMiD: I'm somewhat puzzled why the highest increase in code size did not happen for iriver, but for fmrecorder...
20:43:15dpassen1by pressing up on top
20:43:16preglowTiMiD: you could before as well
20:43:20preglowTiMiD: you just pressed one more time
20:44:27TiMiDamiconn: that's because fmrecorder have buttonbar, clock and so on, so there is more duplicated code
20:45:08amiconnI checked the exact numbers, btw: fmrec: +4464 bytes, h120: +3924, ondiofm: +3756, ondiosp: +3760, player: +2288, rec: +4400
20:45:23 Nick paugh is now known as Caffeinez0r (n=kickback@2001:5c0:8fff:ffff:8000:0:3e03:6822)
20:46:37preglowlinuxstb: and it only stops somtimes, heh
20:46:45linuxstbpreglow: Yep :(.
20:46:50preglowlinuxstb: so it does sound like something is uninitialised
20:47:46linuxstbI can't see anything though.
20:47:55linuxstbAnd neither can gcc.
20:51:23amiconnBtw, I had an idea how to handle compressed firmware images for recorder v1 (and perhaps player some time).
20:51:44preglowis that so hard?
20:52:12amiconnIn fact we already have a big part of what we need, rockbox.bin is ucl-compressed by the build system anyway
20:52:31amiconn...intended for flashing atm, but it can be reused
20:53:11amiconnThen we just need a tiny loader that gets the .ucl "tacked on", then scrambled to be read by the archos loader
20:53:36amiconnWe already have the ucl decompressor as well, as part of our flash bootloader
20:53:42preglowjust have the loaded copy itself to iram and decompress, then skip to entry
20:54:30amiconnThe loader will possibly have to copy the compressed part to upper ram, the decompress into the final place
20:54:44preglowahh, yes
20:54:55amiconnWe can't decompress in-place
20:55:02preglowno, of course you can't
20:55:06preglowgotten anywhere with memcpy yet, btw?
20:55:27amiconnI'm tackling archos memcpy first, for several reasons
20:55:37preglowhmm? hasn't archos already got one?
20:55:48amiconnYes it has
20:55:56amiconnI got it a bit faster for medium sized blocks, but tiny blocks are slower now
20:56:09amiconnI want to test some techniques, also for adding memmove()
20:56:45amiconnmemcpy will be a sub-function of memmove, so it won't take any extra space
20:57:19amiconnThere are basically two ways to implement memmove:
20:58:31amiconn(1) Have a fully optimised forward copy, and a simple backward copy. Use the latter only if dest_addr > src_addr *and* the blocks overlap
20:59:39amiconn(2) Have fully optimised copy routines for both directions, and decide simply by the address relation
21:00:19amiconnI intend to do (2)
21:01:24preglowgood luck, hehe
21:01:35 Quit Moos (Read error: 104 (Connection reset by peer))
21:01:38 Join DrMoos [0] (
21:02:35 Nick DrMoos is now known as Moos (
21:10:02linuxstbSlasheri: Are you around?
21:11:02linuxstbI have a theory about the FLAC bug. I think it may be that when skipping the metadata, I am calling ci->advance_buffer() beyond the amount of data read into the audio buffer.
21:11:42preglowthere are still 16kb iram left for libmad...
21:13:29linuxstbOne odd thing about the dbpoweramp encoded FLACs is that they have about 32KB of padding in the metadata.
21:14:28preglowsame for vorbis, 18kb for wavpack, 11kb for flac, 28kb(!!) for musepack, 25kb for a52
21:14:31preglowthere's some work to do
21:14:45preglowlinuxstb: damn, what the hell is it expecting to have appended there? jpegs?
21:15:09linuxstbdbpoweramp has an odd attitude to metadata. Probably wants to put some id3v2 tags in there.
21:15:44linuxstbI going to change my advance_buffer() call to a read() and see if that helps.
21:16:27preglowa52, flac and wavpack can probably be put in iram in their entirety
21:16:39amiconnThey can't
21:16:53amiconnThe copy routine must stay in dram ;)
21:16:53preglowa52 can, _almost_
21:17:07preglowand wavpack can just be contained in iram
21:17:25linuxstbIt would be impressive to have codecs running in 48KB of RAM.
21:17:44preglowamiconn: btw, i think it's a grand idea to introduce proper bss behaviour for codecs. all codecs are loaded in the mp3 buffer before the track where they're needed, so this will gain us some extra buffer
21:17:51amiconnpreglow: What about musepack? ;)
21:17:56*amiconn hides
21:17:56linuxstbBut I think most of them have buffers in dram as well though.
21:18:07preglowamiconn: yes, i don't even have room for the main struct of musepack in iram
21:18:35amiconnYeah, and th emain struct contains the buffers iirc :(
21:18:36preglowi wonder what the hell the musepack people were smoking when they did that, i'm sure some of the buffer can be used several times
21:18:40amiconnStrange design...
21:19:16preglowamiconn: i tried moving the main buffers into iram and just putting buffer pointers in the struct, but alas, i didn't even have space for the buffers...
21:20:02amiconnWas that before or after increaisng codec iram?
21:20:08preglowafter, i believe
21:20:32preglowdo we have an ICODE_ATTR ?
21:20:49amiconnThat one existed for ages...
21:20:58preglowi never used it
21:20:58 Join miner49er [0] (
21:21:15 Join Musicmad [0] (
21:21:56preglowamiconn: but what do you say about ibss in codecs?
21:22:04Slasherilinuxstb: Hmm, in that case advance_buffer should lock and wait until the buffer has enough data
21:22:43Slasherior in fact it tries to seek and rebuffer then
21:24:33miner49erHi there, I'm going to have another go at building the cross compiler again. I used gcc-3.3.6 previously. Any Ideas? Is it worth using a lower version off gcc?
21:24:50Musicmadhe guys, was messing around with recording. Is it possible to adjust gain and see it visually while/before recording?
21:25:12KohlrabiI built Crosscompiler with gcc-3.3.3
21:25:41amiconnpreglow: In fact we should handle both .bss and .ibss properly. They add up to >64KB for some codecs...
21:26:08SlasheriI use gcc-3.4.4
21:26:16SlasheriAnd binutils-2.16
21:27:30preglowamiconn: agreed
21:27:48preglowhow large is the codec plugin buffer at the moment?
21:27:55amiconn(flac, alac, mpc)
21:29:00amiconnmpc is "top" - ~120 KB .bss, 18 KB .ibss
21:29:13amiconncodec ram is 256KB
21:30:18preglow120kb bss, ahahahha
21:30:31preglowthat'd be the main struct
21:30:39*preglow whipes tears
21:31:02amiconnYes it is except 24 bytes other data
21:32:12preglowi can't believe this is essentially an mp2 decoder
21:33:20amiconnWhen changing the bss handling, we need to keep codec swapping in mind.
21:34:12preglowoh? how does it change anything?
21:34:19preglowwell, it does if the swapper doesn't copy the entire buffer
21:34:30preglowwhich it of course should avoid :V
21:34:54preglowi hate plain binary images
21:35:42preglowlibmusepack uses a _64K_ big read buffer
21:35:44*preglow faints
21:35:58solexxHow do I recreate my wiki-user page?
21:36:38solexxmy username and pw are still accepted, but when using the name on a page, i can only create a "regular" page
21:36:46amiconnThe codec swapper does indeed swap the whole buffer...
21:40:28amiconn...and it uses two buffers each for the dram and iram part. 608KB total
21:41:40SlasheriThe two buffers was used for performance reason, so there is less copying
21:41:51Slasheritypos++ :D
21:42:11amiconnYes. Otherwise it would be necessary to cross-copy byte-wise
21:42:56amiconnNah, not byte-wise, but a few bytes at once (a long or such)
21:43:30SlasheriI tried that and it caused much more delay to the swapping
21:43:49amiconnThis could still be reasonably fast with assembler though, but we would have to retain the plain C method
21:44:04amiconn...e.g. for the ipod port ;)
21:44:08Slasherihehe :D
21:45:01amiconnThe good thing with the simple swap is that it's immune to bss/ibss handling changes
21:47:38amiconnA cross-copy can be made really fast when all buffers are line aligned; the sizes are multiples of 16 anyway
21:48:33amiconnThen it would use burst mode (but memcpy will hopefully use burst soon as well)
21:49:17 Join muesli- [0] (
21:52:40 Join matsl [0] (
21:58:06 Join XavierGr [0] (
22:00:05muesli-re XavierGr
22:00:08linuxstbI don't know why, but I think I've fixed the FLAC bug.
22:00:36linuxstbI replaced the advance_buffer() call with a while loop perform reads of 1 byte. It now seems to work reliably.
22:01:55 Join pangolin [0] (i=pangolin@
22:01:56amiconnTiMiD: Did you copy the old splash code for your gui_splash ?
22:02:03 Quit pangolin (Read error: 104 (Connection reset by peer))
22:03:01 Join ep0ch [0] (n=ep0ch@
22:03:29preglowi just did a little musepack opt
22:03:41ep0chquick question, when in ''screenshot enabled mode', how to do i take a screenshot?
22:03:54linuxstbInsert the USB cable.
22:04:11preglowiriver-xtreme.mpc now plays at 30% boost ratio
22:04:25linuxstbWhat was it before?
22:04:26ep0chhmm, there docs on the screenshot mode?
22:04:34preglowlinuxstb: i didn't check :>
22:04:52preglowbut this SHOULD be a drastic improvement
22:05:06preglowi searched for the place in the code with the biggest number of memory accesses
22:05:39preglowand to my "surprise", this was in the subband filter innerloop
22:05:47ep0chahh cheers. i'm not lazy to seach, honest :)
22:06:13preglowand i could _JUST_ squeeze in two very important buffers
22:07:17preglowoh my...
22:07:58preglowold boost factor was above 90%
22:08:22 Join ender1 [0] (i=ychat@
22:09:38dpassen1thats a phenomenal optimization, thanks and congrat
22:10:07ep0chespecially when it was a 'little' optimisation :)
22:10:08preglowso you think i should commit it? :P
22:10:39ep0chcan you do the same with Vorbis please :)#]
22:10:42dpassen1if you feel like it
22:11:40dpassen1looks like soon, i
22:11:47Musicmadpreglow: had a take on the wma float > integer conversion?
22:11:48dpassen1ill be loading my mpcs on my 120
22:12:09preglowMusicmad: i have no interest in wma
22:12:18TiMiDamiconn: indeed
22:12:43Musicmadpreglow: oh - I though you once said you'd do it for the challenge. Must have been someone else then.
22:12:44preglowMusicmad: but i _MIGHT_ do it one day, to properly teach myself how to use fixed point math
22:12:52ep0chdidn't someone say they were going to take on the DUMB challenge?
22:13:04amiconnTiMiD: Okay, then that still needs some fixing for proper bidi, and an ugly glitch, unless the unicode patch fixes it as well
22:13:05preglowep0ch: that'd be me as well, but i've never got time for it
22:13:21ep0chpreglow: nah someone else said they wanted to have a go at it
22:13:39preglowoh well
22:13:46preglowthat shouldn't be too hard, at least
22:14:39ep0chwould i be right in saying the emac would be very good at doing all the channel mixing in dumb?
22:17:00preglowi'll commit them usepack opt now
22:17:10preglowep0ch: nah, i doubt it
22:17:15amiconnTiMiD: The buttonbar is probably very special, in that a unit might exists that has softbuttons on the remote, but not on the main unit, or vice versa. Or it could have a different number of softbuttons on each...
22:17:23preglowep0ch: if all the channels are mixed in parallel, then yes, but i doubt they are
22:17:52amiconnBut don't code for the future, I'd say. Just keep it so that it's extendable if needed
22:18:20ep0chamiconn: you know since you optimised the peak meter, i get some strange black band at the top end of the peakmeter, i'm trying to get a screenshot for you...
22:18:51 Join XavierGr_ [0] (
22:19:00TiMiDamiconn: It's possible with my patch
22:19:13preglowep0ch: clipping indication
22:19:17XavierGr_wtf is going on with my connection?
22:19:18 Quit ender` (Nick collision from services.)
22:19:21ep0chstupid me!
22:19:21 Nick ender1 is now known as ender` (i=ychat@
22:19:27XavierGr_hi all.
22:19:32***Saving seen data "./dancer.seen"
22:19:39TiMiDyou can have as many buttonbars as you want and you can attach tem to screens
22:19:44TiMiDhi XavierGr_
22:19:48amiconnNice :)
22:20:00TiMiDhuhuhu ;à
22:20:24preglowmusepack must really be fast if you've got tons of ram and a full 64 bit accumulator
22:20:59preglowtons of fast ram, that is
22:21:02dpassen1its the fastest lossy format on x86 i believe
22:21:08ep0chheh i was gonna say fast ram
22:21:11 Quit Musicmad ("Trillian (")
22:21:16preglowdpassen1: as it should be
22:21:45dpassen1foobar did a nice decoding test with tons of different chips and formats and MusePack was the winner
22:22:23ep0chdpassen1: got a link?
22:22:24amiconnpreglow: In fact it might be possible to make the variable shift multiplication faster for many cases. Many cases would only need the emac and reading accect
22:22:28amiconn*accext even
22:22:32dpassen1ep0ch: will look
22:23:20 Quit linuxstb (Connection timed out)
22:23:50ep0chnice, ty
22:24:05 Join linuxstb [0] (
22:24:51dpassen1from a users standpoint a very nice format
22:25:46ep0charent the bitrates quite high for it though?
22:25:56dpassen1depends on what you consider high
22:26:03amiconnpreglow: The fixed shift mul can also be made faster, but that needs integer mode...
22:26:08dpassen1at 140 kbps and up, it tends to be more transparent than any other lossy
22:26:19preglowamiconn: oh? how?
22:26:28dpassen1MPC −−standard tends to save me about 20 kbps over LAME MP3 −−aps
22:26:34amiconnHmm, in fact my idea would need switching between integer and fractional mode...
22:26:48preglowwell, that'd work okay
22:26:48ep0chshame i've now encoded everything with vorbis@q4
22:26:52preglowthe shift itself takes just two cycles
22:27:04linuxstbSlasheri: I'm convinced there is a subtle bug in the rebuffer_and_seek function, but I can't see it. By using read() instead of advance_buffer(), previously troublesome FLAC files play perfectly.
22:27:19amiconnpreglow: The fixed mult shifts right by 14. The acc is 48 bit, so we don't need separate calculation of low & high...
22:27:20dpassen1not really, vorbis has gapless, replaygain, and good tag support
22:27:31preglowamiconn: oooh, i see
22:27:53ep0chdpassen1: mpc is gapless or not?
22:27:59preglowdpassen1: and last but not least, you CAN SEEK
22:28:09preglowep0ch: yes, through hacks, it seems
22:28:17dpassen1preglow: forgot that, but i tend to speak more through a users standpoint
22:28:28preglowdpassen1: and users suddenly don't seek anymore? :P
22:28:36ep0chimho anything that adds a gap is useless
22:28:47dpassen1doesn't need to be sample accurate or incredibly fast on modern processors
22:29:12dpassen1i can imagine scaling it down to the processor in the 1x0s is very different
22:29:33*amiconn still prefers plain old mp3. Plays almost anywhere, and is quite good when created with lame
22:30:13dpassen1can't argue with that, i have nearly 14 GiB of LAME -aps encodes
22:30:16preglowi prefer free formats whenever i can
22:30:17ep0chyeah you cant beat the portability of mp3
22:30:18markunamiconn: phaedrus961 fixed bidi in the unicode patch I believe
22:30:56amiconnmarkun: That's good for sure, but I'm afraid we now have to wait some time before unicode can be committed :(
22:31:42markunYes, I'm afraid of that too. Maybe if you would make the bootimages self-extracting.. ;)
22:31:53preglowamiconn: btw, the accumulator is 40 bit in integer mode
22:32:01preglowamiconn: not 48, the upper extension word is just a sign extension
22:32:31amiconnMeh :(
22:33:08Slasherilinuxstb: Hmm, could you send one problematic file to me so I could try that later?
22:33:12 Join einhirn [0] (
22:33:20amiconnUsually I do read datasheets thoroughly. Overlooked that.
22:33:28ep0chwill rockbox have a plugin to flash the firmware for iriver? i hate going into stock firmware to flash...
22:33:34preglowamiconn: tickled the back of my mind, so i checked it
22:33:52preglowep0ch: yes, we probably will
22:34:21Slasherilinuxstb: ok, thanks :)
22:34:35ep0chit will be needed if we ever have the whole of rockbox as firmware :)
22:35:03phaedrus961amiconn: what is the problem with bidi and the splash function?
22:35:36linuxstbSlasheri: The issue that worries me is that it sometimes works, sometimes doesn't. But the logf seems to point to the rebuffer_and_seek function being called each time it fails.
22:35:51amiconnThe problem (in cvs code) is that the splash function draws the lines word by word, not the whole line at once, so the bidi algorithm can't work correctly
22:35:52 Quit XavierGr (Read error: 110 (Connection timed out))
22:36:48Slasherilinuxstb: Hmm, good. Then the problem is probably with that function. I will take a look soon, cu ->
22:37:23phaedrus961ahh ok, then it hasn't been fixed by the unicode patch
22:38:37solexxhi again. anybody got a solution to my wiki problem? (wrote about it an hour ago)
22:40:02preglowi've got no idea
22:40:07preglowi just plain and simple registered again
22:40:55solexxcan I delete my account? I am still registered, since I can still login
22:42:41solexxI forgot, who is administering the site again?
22:42:57preglowbut what about someone implementing this compressed rockbox image thing so we can have unicode as well? :P
22:43:41amiconnWe will have it, for sure.
22:44:41solexxAm I right that unicode support won't have any effect on the filesystem?
22:44:45solexx(reading and writing)
22:45:05*solexx is talking about filenames
22:45:16amiconnI think it wouldn't be a good idea to incorporate two major changes at once. It might make debugging very hard
22:46:28amiconnSo imho we should just wait with unicode commit until the gui conversion is done.
22:46:53amiconnThat doesn't mean to stop unicode development...
22:47:00phaedrus961solexx: not exactly, the unicode patch writes filenames in proper unicode
22:47:16phaedrus961the current code only writes proper unicode for latin1 names
22:47:50 Quit XavierGr_ (Read error: 110 (Connection timed out))
22:53:29ep0cham i right in saying the peakmeter values are calculated before replaygain applies gain?
22:54:49ep0chi.e. before the dsp
22:55:18 Quit Bagder ("Off to search for that connect-resetting peer guy!")
22:55:19linuxstbpreglow: I would be happy for you to enable your EMAC changes to the FLAC decoder. I don't think we need to wait for seeking.
22:55:52preglowaight, gimme half an hour
23:00:07ep0chmeh, ignore me i managed to test the only album i have that doesn't have replaygain tags
23:01:01 Join muesli__ [0] (
23:16:16 Part ep0ch
23:24:01 Quit muesli- (Read error: 113 (No route to host))
23:29:33 Join Kohlriba [0] (
23:31:13*amiconn found a bug in the Ondio builds
23:32:00miner49erWow! I think GCC has finally built! it took about 2 hours, but I think it's worked! Is there any way I can check if it's built, I mean it doesn't say Erroor anyway, it ended with nothing to do for install
23:32:48 Quit Kohlrabi (Nick collision from services.)
23:32:49 Nick Kohlriba is now known as Kohlrabi (
23:35:30TiMiDhow can I update just a directory with CVS ?
23:37:57amiconnFound the problem why BUTTON_OFF didn't stop playback from the file browser on iriver and Ondio as intended
23:39:40 Quit ender` (Read error: 104 (Connection reset by peer))
23:41:31 Join linuxstb_ [0] (
23:42:28linuxstb_TiMiD: I'm not sure what you mean. If you do "cvs update" then that will update everything in the current directory (and subdirectories). Is that not what you want to do?
23:44:55TiMiDwhat I want to do is to update just the apps dir and all the subdirs
23:45:32linuxstb_Then just do "cd apps; cvs update"
23:46:09miner49erI'm trying to build the X11 sim, can anyone help me out?
23:46:21TiMiDwill it be aware that it must be updated on app/ on the server ?
23:46:30TiMiDit parses the path I suppose
23:46:43linuxstb_TiMiD: Yes - every directory has a "CVS" subdirectory containing that information.
23:47:11linuxstb_e.g. you could move the apps/codecs directory anywhere on your hard drive, and still update that codecs directory doing "cvs update" from inside it.
23:47:21TiMiDok !
23:47:29TiMiDI understand know ;)
23:47:39TiMiDthanks for the lesson :)
23:48:32linuxstb_I also find it very useful to do "cvs diff -ub file.c" before I commit changes to that file. That will show you what you are about to commit.
23:49:15amiconnWhat does the -b do?
23:49:37amiconnI'm doing similar, usually a diff -u across the whole working copy
23:50:07amiconncvs diff -u even
23:50:56linuxstb_I'm not sure I should be using it - it ignores changes in whitespace.
23:51:04linuxstb_But I remember badger mentioning it once.
23:51:27linuxstb_In fact, I normally do a variety of diffs - sometimes just a simple "cvs diff" (no options) and somtimes -u3
23:51:39miner49erI'm trying to build the basic firware, giving the sim a break. But I'm getting ***No rule to make target 'gui/buttonbar.c' ...can anyone pleae help me?
23:52:06TiMiDgrrr encoding :(
23:52:24amiconnminer49er: You need to reconfigure
23:52:32preglowlinuxstb: done
23:52:51miner49eramiconn: you mean (in build dir) ../tools/configure ?
23:53:20TiMiDI will remove the é from my name :/
23:53:47miner49erjust did that, and i'm getting the same cobblers back
23:54:09linuxstb_miner49er: Do you have the apps/gui/directory in your source tree?
23:54:19linuxstb_I mean apps/gui/ directory
23:54:39miner49erhang on, i'll check
23:55:38miner49erlinuxstb: It appears not, just a load of .c and .h files
23:55:52linuxstb_In which case do "cvs update -d" from the apps/ directory. The -d option means to get new directories from the repository.
23:56:28miner49erwell, I just downloaded the lates source file - I havn't delved into cvs yet. Is it possible to do this now?
23:56:42miner49er...or shall I start from a blank source dir?
23:58:09linuxstb_The gui/ was missing from the source files.
23:58:26miner49erRight, I'll just get to grips with CVS then, cheers!
23:58:31linuxstb_I added it earlier today - is there a "bleeding edge" source archive?
23:58:46linuxstb_But yes, CVS will be easier.
23:59:00miner49erI'd prefer to learn how to use CVS really. it's A Good Thing

Previous day | Next day