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 2005-11-01

00:00:25 Quit Kohlriba ("Leaving")
00:01:00elinenbeamiconn: the unicode patch is very nice... it would be great if that was merged into CVS one day.
00:02:27preglowlinuxstb: i can't even find where to control the bitrate in this thing
00:03:09amiconnelinenbe: Be patient... first TiMiD has to finish his remote work
00:03:33linuxstbpreglow: I guess you've installed iTunes?
00:04:26preglowlinuxstb: aye
00:04:27 Quit Maxime (Read error: 104 (Connection reset by peer))
00:04:54 Join Maxime [0] (
00:06:40linuxstbI don't know about the Windows version, but on the Mac, it's the application menu, then preferences, then advanced, then importing.
00:07:18TiMiDamiconn: about the code size decrease, I'm becoming skeptical :) (well once the work finished, we will return around the initial level, but not as big as I expected : the menu's new implementation reduced it by 300bytes ... very poor )
00:07:22_FireFly_it seams that f2_screen, f3_screen and handle_usb in wps.h are not used can they be removed ??
00:07:26preglowlinuxstb: yeah, i found it
00:07:31preglowlinuxstb: didn't notice the tabs
00:07:35 Quit matsl (Remote closed the connection)
00:08:06preglowlinuxstb: is vbr aac normal?
00:08:44linuxstbI don't think iTunes does VBR. But Nero does.
00:09:25linuxstbJust noticed the VBR checkbox - I guess it does it now.
00:11:19preglowhrmph, MUL_R is used not only for sbr
00:11:29preglowbut most extensively in sbr, it seems
00:15:48*preglow cuddles MUL_F
00:15:57preglowi love it when the fixed point formats line up with what the emac unit uses
00:17:38linuxstbIs that good news then?
00:18:28preglowlet's hope for some long accumulator chains
00:18:35preglowalthough i'm pretty certain they aren't there
00:18:51preglowaac doesn't use any good old-fashioned fir filtering
00:18:55preglowthat i know of, at least
00:19:53preglowbad news is that the REAL_PRECISION is used in a couple of places
00:20:00preglowwe might need to do a proper 64 bit multiply for that
00:21:28linuxstbIn case you're wondering, my choice of what is in IRAM is pretty random. So feel free to change it if you think it can be better used elsewhere.
00:21:34amiconnThe full 64*64->64 ?
00:23:04preglowamiconn: no, musepack style
00:23:37preglowbut ok, i'm doing a couple of tests right now
00:23:38amiconnThat's not too bad. Variable shift as well?
00:24:37preglowamiconn: nope, fixed shift
00:24:51preglowamiconn: but the majority of all muls will be emac
00:25:04preglowdoes anyone know if the "cc" clobber flag is really necessary?
00:25:37amiconnI'm not 100% sure, but I never used it and didn't run into problems
00:30:04*preglow spots a two-space indent level and looks at the culprit :P
00:30:47_FireFly_it seams that the functions f2_screen, f3_screen, refresh_wps and handle_usb in wps.h are not used can they be removed ??
00:31:59LinusN_FireFly_: yes
00:32:22_FireFly_i'm trying to make a wps-widget :)
00:34:52linuxstbpreglow: Fixed in CVS. I'm trying to comply...
00:35:43preglowi used to to two-level indenting myself when i used pascal
00:35:51preglowwhich is quite luckily very long ago
00:36:25linuxstbMaybe that's where I picked up the habit from - I used pascal for a long time.
00:36:56 Join len0x_ [0] (
00:37:13elinenbe_FireFly_: a widget that does what?
00:37:59 Quit Maxime (Read error: 104 (Connection reset by peer))
00:38:18 Quit _DangerousDan ("Miranda IM! Smaller, Faster, Easier.")
00:38:19 Join Maxime [0] (
00:38:26_FireFly_draws the wps :)
00:38:31preglowreplacing the math function alone isn't going to save aac, it seems
00:38:44preglowargh!!!! ice!
00:39:37len0x_what's taking the most time in AAC decoding? iDCT?
00:39:46preglowi have no idea
00:39:48_FireFly_elinenbe: i try to adapt the current wps code to the new multi-screen support
00:40:13preglowlen0x_: not all words starting with an 'i' has to have a lowercase i, you know, it's called an IMDCT ;)
00:42:03len0x_back in the days it was just iDTC :)
00:42:15len0x_iDCT :)
00:42:23 Join RotAtoR [0] (
00:42:45len0x_anyway - does it use asm ?
00:42:54preglowlen0x_: nope
00:43:06preglowlen0x_: everything but trivial multiply functions are c
00:43:21Mode"#rockbox +o LinusN " by ChanServ (ChanServ@services.)
00:43:34Kick(#rockbox len0x :LinusN) by LinusN!
00:44:11Mode"#rockbox -o LinusN " by LinusN (
00:46:03 Quit len0x_ ("Trillian (")
00:47:39 Join len0x [0] (
00:48:22preglowwhat the hell is it about libfaad that keeps provoking these ICEs
00:49:21linuxstbDid the version in CVS compile for you?
00:50:11preglowit worked just fine with a replaced MUL_F
00:50:20preglowi've now replaced MUL_C, and it ICEs in tns.c
00:50:48preglowtrying to replace functions with macros
00:51:03preglownope, same error
00:51:49linuxstbOne of the problems I had was the faad_getbits() function. It was originally inline, but that caused ICEs. So I had to make it a normal function.
00:52:15linuxstbMaybe it's worth trying gcc4
00:53:06preglowlinuxstb: but now i made the mul functions macros, and it still didn't help. making them ordinary functions are out of the question
00:53:41 Join einhirn [0] (
00:53:47preglowis tns.c necessary for lc operation?
00:54:16linuxstbI have no idea.
00:55:09preglowi'm using gcc 3.4.1, however, could you test out the file with your gcc if it's newer?
00:55:39linuxstbSure. But I'm not confident it will be any better.
00:55:47linuxstbI'm using 3.4.4
00:55:50 Part len0x
00:57:46 Quit tvelocity ("Leaving")
00:59:05linuxstbICE :(
00:59:57preglowwe've got heaps of codecs with no ICEs, and now suddenly, tons of them in one codec
01:02:50 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
01:05:20amiconnI already got memcpy() 3.5x..4x as fast as the C version for most alignment combos. That's still without using burst mode... :-)
01:08:00MoosSeems promising :)
01:08:54amiconnOtoh, this might become a pretty large beast
01:09:19preglowabout how large?
01:10:16amiconn242 bytes already. Using burst mode optimised for various alignments might easily quadruple this.
01:10:25amiconnmemmove will double the size again
01:11:01preglowwell, as long as the speedup is goog enough
01:11:06preglowgood, even
01:11:22preglowbut it might be a tad too large for putting in iram
01:19:59 Quit _FireFly_ (Read error: 110 (Connection timed out))
01:21:04***Saving seen data "./dancer.seen"
01:28:29preglowstill not realtime, hrmph
01:34:48linuxstbDid you solve the ICE problem?
01:35:26 Nick paugh is now known as AliasCoffee (n=kickback@2001:5c0:8fff:ffff:8000:0:3e03:6822)
01:38:05preglowi'm just ignoring that particular mul function for now
01:38:20amiconnLinusN: Is there a reason why you didn't enable SDRAM page mode in the bootloader?
01:40:53LinusNamiconn: instead of burst-only page mode?
01:41:50 Join ashridah [0] (
01:43:49amiconnAfaiu, this isn't really 'instead of'. If PM isn't set, the coldfire uses page mode for bursts only. If it's set, the coldfire uses page mode for both burst and normal accesses
01:44:42amiconn(when appropriate)
01:45:08LinusNamiconn: iirc, i interpreted it as "instead of" for some reason, but now that i look at it, i seem to have been wrong
01:45:12amiconnThe benefits might be small, because the coldfire doesn't have warp mode & ras down
01:45:27LinusNand the pipeline is rather lame
01:45:52LinusNstill, i don't see a penalty
01:46:09preglowhow is the pipeline lame?
01:46:10LinusNbut there has to be one, otherwise it wouldn't have been an option
01:49:19amiconnMCF5249UM.pdf, section "The use of continuous page mode is recommended because it can provide a slight performance increase."
01:50:33 Quit linuxstb ("CGI:IRC")
01:50:35LinusNpreglow: it's short, and doesn't group data access and instruction fetches very well, so it makes bad use of the burst controller
01:51:11LinusNadmittedly, it's not all about the pipeline, it's more about the memory controller
01:51:23preglowexactly how short is it? i've tried to find some info about the pipeline, but i've failed miserably
01:54:13 Join Mxm`Pas`Bien [0] (
01:54:13 Quit Maxime (Read error: 104 (Connection reset by peer))
01:55:05 Nick linuxstb_ is now known as linuxstb (
01:56:13 Quit Moos ("Glory to Rockbox")
01:56:31 Join Moos [0] (
01:59:02amiconnWow... memory accesses on coldfire are pathetic. I did some calculations based on my memcpy() timing
01:59:26preglowwe know that...
01:59:46amiconnA loop that would take 5 cpu clocks per pass according to the instruction timing, but copies one long from dram to dram takes 43 clocks (!!)
02:00:20preglowtried enabling page mode to see if it gets better?
02:00:51amiconnPage mode wouldn't help in this case. Source & destination are in different pages
02:01:33pregloware you sure the memory timing is as tight as it can be?
02:02:06 Quit Moos ("Glory to Rockbox")
02:02:18 Join Moos [0] (
02:02:28LinusNyes it is
02:03:55preglowoh well
02:07:59LinusNunfortunately, the sdram controller clock is always 1/2 the cpu clock
02:08:07LinusNeven at 11mhz
02:08:22LinusNa major suckfactor
02:08:49preglowman, aac is pretty memory heavy
02:18:02preglowwe could have half a meg of iram for this and still not have enough
02:20:18LinusNgotta sleep, nite all
02:20:35 Part LinusN
02:22:26preglowno matter what i stuff into iram, it's still dog slow
02:31:12 Quit Slasher (
02:39:06 Quit Moos ("Glory to Rockbox")
02:39:53 Join Naked [0] (
02:43:33NJoinSlasher [0] (
02:46:45 Quit AliasCoffee ("Leaving")
02:50:32 Quit Hadaka (Read error: 111 (Connection refused))
02:51:11 Nick Naked is now known as Hadaka (
03:11:53 Quit cYmen ("zZz")
03:21:08***Saving seen data "./dancer.seen"
03:28:58 Quit Mxm`Pas`Bien (Read error: 104 (Connection reset by peer))
03:29:26 Join bagawk [0] (n=lee@unaffiliated/bagawk)
03:32:48 Join Maxime [0] (
03:51:14 Join Mxm`Pas`Bien [0] (
03:51:14 Quit Maxime (Read error: 104 (Connection reset by peer))
03:55:22 Join Lost-ash [0] (
03:55:24 Quit ashridah (Nick collision from services.)
03:55:43 Nick Lost-ash is now known as ashridah (
04:38:06 Join Paul_The_Nerd [0] (
05:01:13bagawkBagder, good day :-)
05:21:13***Saving seen data "./dancer.seen"
05:41:46 Join Maxime [0] (
05:41:46 Quit Mxm`Pas`Bien (Read error: 104 (Connection reset by peer))
05:48:04 Quit Paul_The_Nerd ("Chatzilla 0.9.68a [Firefox 1.0.7/20050915]")
06:19:22 Join Serotta_b [0] (
06:19:22 Quit Maxime (Read error: 104 (Connection reset by peer))
06:19:30 Join Maxime [0] (
06:20:34 Part Serotta_b
06:39:33 Quit bagawk (""umount /dev/brain"")
06:40:09 Join matsl [0] (
06:49:31 Quit RotAtoR ()
07:21:17***Saving seen data "./dancer.seen"
07:35:48 Join Mxm`Pas`Bien [0] (
07:35:54 Quit Maxime (Read error: 104 (Connection reset by peer))
07:36:56 Nick Mxm`Pas`Bien is now known as Maxime (
07:57:27 Quit matsl (Remote closed the connection)
08:05:55 Quit lostlogic ("Going to the moon")
08:16:39 Join solexx_ [0] (
08:28:09 Quit solexx (Connection timed out)
08:33:50 Join webguest82 [0] (
08:34:50webguest82There is that wonder of some piece. Do you have person who reply?
08:35:44webguest82Can you find out iRiver H300's firmware source?
08:35:49webguest82Do you have person who answer?
08:36:47ashridahthe source code to the iriver firmware? we won't ever get that
08:37:39webguest82Too ..
08:38:15webguest82Is not it hard to manufacture one after another?
08:38:34ashridahwe can disassemble it.
08:38:48ashridahsomeone probably already has to a degree.
08:39:14ashridahthe only thing impeeding H300 development at the moment is really manpower
08:39:31ashridahthe people who have the tools who can work on it are very busy with work atm.
08:40:10webguest82What work is ashridah doing?
08:40:55ashridahnot much. i pretty much test rockbox against my H140 whenever i get time
08:41:28webguest82Are there much that know about H300?
08:41:43 Join einhirn [0] (
08:42:24ashridahmost of the problems facing rockbox on H300 involve tracing the hardware and working with the unit with a BDM
08:42:44ashridahas i say, the rockbox developer with the tools to do that is currently busy.
08:42:47thegeekyou sound like an alien webguest82
08:43:53webguest82thegeek: So.
08:44:21ashridahthegeek: i'm guessing he's using a translator like babelfish. wasn't there someone else who found a better one?
08:44:29webguest82I am H300 user who live in South Korea.
08:45:06webguest82Used babelfish before, but unuse now.
08:45:19webguest82Babelfish is miserable.
08:45:35ashridahit doesn't seem to have improved too much
08:46:00webguest82Speech which do now may understand.
08:46:17webguest82Is using program of HaanGuide.
08:47:19webguest82Where does ashridah live?
08:47:25ashridahanyway, as i was saying, the main obstacle with H300 and rockbox is free time for the developers who are working on it.
08:47:31ashridahI'm in Australia.
08:48:34webguest82Live in the good country.
08:48:56webguest82The South Korea is complicated and dizzy.
08:48:57thegeekashridah : yeah, I guessed that
08:49:10thegeekactually webguest82
08:49:14thegeekit was not that bad
08:49:24thegeekI've seen much worse results from babelfish
08:49:38thegeekit was just wierd enough to sound like an alien;)
08:50:11webguest82Where does thegeek live?
08:51:01thegeekway up north
08:51:02webguest82Does not the South Korea have a person who work in RockBox?
08:51:20webguest82Norway is the great country.
08:51:21thegeekI dont think so
08:51:26thegeekyep, it's nice:)
08:52:28webguest82Do you know meaning of 'OTL'?
08:52:47thegeeknot really;)
08:53:25webguest82Do not I know?
08:53:37webguest82Do not you know?
08:53:49webguest82(was mistake)
08:54:24thegeekI do not know what it means;)
08:54:28webguest82'OTL' is Emoticon that express image that a person is miscarrying.
08:54:49thegeekah, I see, it means you dont understand ?
08:54:52webguest82Various Emoticons are used in the South Korea.
08:55:35webguest82The South Korea is that is the bad country.
08:55:54thegeekbad country?
08:55:57thegeekyou dont like it there?
08:56:08webguest82How does the South Korea think about the country?
08:57:08thegeekotl ;)
08:58:37webguest82'OTL' is that express that a person miscarries.
08:59:20webguest82O - Head
08:59:20webguest82T - Trunk
08:59:20DBUGEnqueued KICK webguest82
08:59:20webguest82L - Leg
09:00:16webguest82Even if the South Korea may go first in technology, the other is bad seldom.
09:01:11 Quit goa ("Client suicide")
09:02:03webguest82By the way, what is BDM Iraneunge?
09:03:48webguest82Oh sorry
09:03:49webguest82What is 'BDM'?
09:04:49 Quit webguest82 ("CGI:IRC (EOF)")
09:05:12 Join webguest82 [0] (
09:05:41ashridahwebguest82: 'BDM' is a special debugging interface for embedded hardware
09:06:12ashridahmany embedded processors such as the motorola coldfire in the iriver devices have special debugging interfaces, the BDM hooks into those (amoungst other things)
09:06:15 Join goa [0] (
09:06:19webguest82Then, is work about H300 childhood yet?
09:09:18ashridahi'm not sure i understood that question, say it another way?
09:11:09webguest82I am sorry.
09:11:13webguest82Is work about H300 startup yet?
09:12:45ashridahit's started
09:12:48ashridahit's just not progressing very fast
09:13:24webguest82Is ashridah busy?
09:14:18ashridahi'm not a developer, i just test on the H140
09:14:27ashridahbut atm i've got exams
09:15:10webguest82I want to do small work.
09:17:04webguest82Can I do if do well C language?
09:17:56ashridahmost of rockbox is in C, and development can be done without extensive hardware specific knowledge. You may be better off asking questions on the mailing list tho. I'm not a rockbox developer as such, just an occasional tester
09:18:30webguest82ha ha
09:20:41webguest82Seem to be many as know still.
09:21:19***Saving seen data "./dancer.seen"
09:22:57 Quit webguest82 ("CGI:IRC (EOF)")
09:23:00 Join webguest82 [0] (
09:23:34webguest82Site structure of rock box is difficult.
09:25:49ashridahit contains a lot of technical information, it might be difficult to read when translated, yeah
09:26:04ashridahunfortunately, it's unlikely that there's much that can be done, i don't believe we really have any developers who speak korean
09:26:41webguest82Is no there a person to do Korean?
09:27:22 Join ender` [0] (i=ychat@
09:28:36webguest82May I how about if I do such work? :)
09:29:57ashridahit's possible. I don't know how much support rockbox's interface code has for korean. You'd definently need to join the mailing list and discuss it, but that might be difficult. the main problem i have is that i'm not a rockbox developer, i can't really say what needs to be done
09:35:07B4gderthere's a unicode patch pending that should enable korean and most other languages too
09:36:02webguest82English learns.
09:36:06amiconnOf course someone would need to create the korean translation
09:38:07webguest82English - > Hangul translation possible.
09:41:35webguest82It does not well that speak in English, but it does well that translate from English into Hangul.
09:47:03webguest82Thank that give help variously today.
09:49:00 Quit XavierGr (Read error: 110 (Connection timed out))
09:49:26 Quit webguest82 ("CGI:IRC (EOF)")
09:57:04markunamiconn: I have a korean translation here.
09:57:47markunand a japansese one
10:03:30B4gderand I have a cup of coffee!
10:08:42linuxstbB4gder: Which version of gcc does the server use to generate the H1x0 simulator auto builds?
10:09:04B4gdergcc (GCC) 4.0.2 (Debian 4.0.2-2)
10:09:14 Quit ender` (Read error: 104 (Connection reset by peer))
10:09:32 Join DangerousDan [0] (
10:09:45linuxstbThat explains the current warnings then. I'm not getting them with 3.3.5
10:10:08B4gderactually when I think about it, the configure outputs the gcc version
10:10:24B4gderbut yes, gcc4 is a lot more picky
10:10:27linuxstbThat's true. Very useful.
10:12:21B4gderthen the gcc4 builds even inhibits some warnings
10:12:29B4gderusing -Wno-pointer-sign
10:14:59linuxstbI don't want to commit a patch without testing it, and I don't have gcc4. But I guess it's about time I upgraded (I'm on Debian).
10:15:32B4gderit seems to work rather well
10:15:36ashridahlinuxstb: the gcc 4 transition's going to start merging into testing soon
10:15:56B4gderI use gcc4 quite a lot
10:16:04linuxstbI'm running unstable - so I'm guessing I can just do "apt-get install gcc-4.0"
10:16:17ashridahlinuxstb: it should already be the default compiler
10:16:35linuxstbIt isn't for me. But I don't run "upgrade" very often.
10:16:39ashridahif you've been dist-upgrading, the compiler itself will be all you're missing, almost everything will have been compiled for it
10:17:01linuxstbI generally just upgrade packages when I need to.
10:17:49B4gdergcc4,, openoffice2 tiny little updates... :-)
10:20:39 Join _FireFly_ [0] (
10:22:32linuxstbI've just installed gcc-4.0, and nothing else was upgraded. So I guess I was already most of the way there.
10:23:02B4gderrerun configure then before you build the sim, or you'll get a bazillion extra bonus warnings
10:23:29linuxstbMmm. /usr/bin/gcc is still a symbolic link to gcc-3.3. Is there a "correct" way to change that?
10:24:14_FireFly_distri ??
10:24:58linuxstbdistri ?
10:25:03B4gdermine is a link to gcc-4.0
10:25:10B4gderbut I don't know exactly what made it so
10:25:33_FireFly_which distribution(debian, fc3, gentoo, ...) do you use
10:25:49linuxstb_FireFly_: Debian unstable.
10:26:45linuxstbOh well, just changed the link manually. I'll soon find out if that breaks anything.
10:28:54ashridahlinuxstb: the 'gcc' package should deal with that
10:29:26ashridahif you'd just installed gcc-4.0 or something, it wouldn't set the default
10:30:09ashridah$ dpkg -S /usr/bin/gcc
10:30:10ashridahgcc: /usr/bin/gcc
10:31:40linuxstbashridah: Thanks. I just did an "apt-get install gcc" and that updated the symbolic link.
10:32:21CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
10:32:21*ashridah nods
10:52:32linuxstbOK, we should have a nice clean build table again now.
10:58:45 Join muesli_- [0] (
11:17:34amiconngcc 4 is already default in debian testing
11:18:21 Join tvelocity [0] (
11:21:21***Saving seen data "./dancer.seen"
11:31:46ashridahamiconn: interesting, didn't think they'd started the c++ abi push yet.
11:49:55 Quit muesli_- (Read error: 110 (Connection timed out))
12:06:17 Quit DangerousDan (Read error: 104 (Connection reset by peer))
12:11:24 Join DangerousDan [0] (
12:22:41 Join cYmen [0] (
12:27:53 Join cYmen_ [0] (
12:31:46 Join muesli_- [0] (
12:32:27 Nick solexx_ is now known as solexx (
12:33:06 Join cYmen__ [0] (
12:35:21 Join Kohlrabi [0] (
12:38:15 Join cYmen___ [0] (
12:38:15***Alert Mode level 1
12:38:15DBUGEnqueued KICK cYmen
12:38:15DBUGEnqueued KICK cYmen_
12:38:15***Alert Mode level 2
12:38:15DBUGEnqueued KICK cYmen__
12:38:15DBUGEnqueued KICK cYmen___
12:38:15***Alert Mode level 3
12:40:03 Join Moos [0] (
12:40:18MoosHello guys
12:40:21 Quit cYmen (Connection timed out)
12:43:18 Quit Moos (Read error: 104 (Connection reset by peer))
12:43:27 Join Moos [0] (
12:43:34 Join cYmen [0] (
12:43:34***Alert Mode level 4
12:43:34***Alert Mode level 5
12:43:34DBUGEnqueued KICK cYmen
12:43:34***Alert Mode level 6
12:43:34***Alert Mode level 7
12:43:34***Alert Mode level 8
12:44:13MoosSlasher: Hi, are you around? I encountered one odd playback bug :(
12:45:36 Quit cYmen_ (Connection timed out)
12:45:44Mooswhen playing one song, and wait for filling bufering
12:46:36Mooswhen it's done, and I skip back for replay song before his end, the player freeze in spining down
12:46:51Mooseasy to reproduce
12:47:03Moosanyone here can reproduce it?
12:47:47SlasherMoos: Hmm, i will try that
12:48:19MoosHi: since your last playback change this week end
12:48:59Slashergood, i can reproduce it
12:48:59 Quit Moos (Read error: 104 (Connection reset by peer))
12:49:00 Join DrMoos [0] (
12:49:24 Nick DrMoos is now known as Moos (
12:49:24SlasherDrMoos: i was able to reproduce it - fix coming soon
12:49:36Moosoki thanks
12:50:41Maxime(ah, one detail, when song is Paused, and you start another, if crossfade enabled, you have the 'end' of the song before starting the other.. starting the other without finishing the paused song will be great, no?)
12:50:46 Quit cYmen__ (Success)
12:51:10SlasherMaxime: Hmm, i can fix that too
12:51:29Maximebecause it's "surprizing" the fadeout of the other song ..
12:52:25Maximeand I always have a bug
12:52:44MaximeI have to start player, play, stop, shutdown, power on again and play again to have sound..
12:53:03Maximethe first start, i have no sound
12:53:06Maximeand the player "lags"
12:53:19Maximeto shutdown, it reacts 20-30s after the button press
12:53:35***Alert Mode OFF
12:54:10SlasherHmm, that's weird :/
12:54:13Maximeit(s strange, it seems i'm the only with this problem
12:54:35Maximeand sometimes the player hangs, i'm forced to reset it
12:54:39SlasherIt's probably hardware related
12:54:50Maximei've tried deleting the whole rockbox folder, resetting config and so on
12:54:54Maximebut nothing changes
12:55:02Maximeand the original firmware haven't this issue
12:56:02Maximebut it's a very annoying issue
12:56:57 Quit cYmen___ (Connection timed out)
12:57:23Slasherunfortunately it's hard for us to fix it because no any developer has encountered it
12:57:36Maximei know
12:57:51Maximebut I have no idea of how to "isolate" this issue..
12:58:37SlasherI you want to try, you could try to change some hardware initialization code from the rockbox source. (Something audio and i2c related for example)
12:58:53Maximebut what(s strange
12:58:59Maximeis that it's one time on two
12:59:06Maxime1. no sound, 2. sound
12:59:32Maximei'll try one thing
12:59:50Slasherbtw, if you use rolo to reload rockbox, do you get sound then=?
12:59:57Maximei'll try
12:59:58SlasherOr do you have to power it completely off?
13:01:52Maximei dont understand
13:01:59Maximeit's when the player is a long time powered off
13:02:16Maximebut .. can't explain this ..
13:02:23Maximecoz no RTC or else..
13:02:27Maximeeating time, bbl
13:06:19 Join muesli- [0] (
13:09:12 Join amiconn_ [0] (
13:14:08 Quit muesli_- (Read error: 104 (Connection reset by peer))
13:16:27Maximesee, the player was powered off when I was eating, and now, no sound
13:16:58 Join LinusN [0] (
13:17:03Maximewhen No sound, when I restart rockbox using RoLo, don't have sound
13:17:13LinusNMaxime: which bootloader version do you have?
13:17:24Maxime2 i think
13:17:31LinusNthat's your problem
13:17:39MaximeIt's a known problem?
13:17:57Maximewhen player was powered of a few minutes, no sound, the restart, sound again
13:18:09LinusNhowever, it should be fixable in rockbox as well
13:18:11linuxstbJust out of interest, what was the fix?
13:18:33 Quit muesli- (Read error: 104 (Connection reset by peer))
13:18:47LinusNthe uda1380 isn't properly reset, so it doesn't reply to i2c commands
13:19:21Maximeso i'm going to upgrade bootloader then..
13:19:33LinusNworth a try
13:19:55Maximebecause this was very annoying lol
13:20:03preglowwhy is the uda init in the bootloader at all?
13:20:18preglowwouldn't it be better if it touched as little as possible?
13:20:34LinusNpreglow: yes, rockbox should reset it as well
13:21:01LinusNthe bootloader resets it to get rid of an annoying "hum" during boot
13:21:25***Saving seen data "./dancer.seen"
13:21:37LinusNand each reset produces a slight "pop", so resetting it twice might be annoying too
13:22:06SlasherHmm, i wonder if we could prevent those reset pops too
13:22:08_FireFly_would be a mute before resetting help
13:22:36LinusNwe can't mute before reset, since it doesn't respond to i2c commands before reset
13:22:49linuxstbpreglow: Any progress with AAC?
13:23:23preglownone at all
13:23:42preglowi'd absolutely love to run some memory access analyzer on this beast
13:23:55preglowreplacing the fixed point functions with emac equivalents didn't help much
13:24:12preglowso it's obvious it's the memory accesses that are bogging stuff down
13:24:15linuxstbI think the memory is the problem - it's using too much slow DRAM
13:24:19preglowand faad most certainly seems to use much of it...
13:24:27linuxstbhehe. I think we agree.
13:24:32preglowi can perhaps fit one or two important buffers in iram
13:24:37preglowthe rest, no way
13:24:57 Join Mxm`Pas`Bien [0] (
13:24:58preglowthe fft/mdct does a ton of memory accesses, so might be beneficial to stuff something around that in iram
13:25:02preglowproblem is, everything's so big..
13:25:05linuxstbDo you think some of the buffers could be merged and therefore share the same IRAM?
13:25:14preglowi have no idea yet
13:25:24preglowi'd need to understand the code to know
13:25:36preglowand right now i haven't even superficially looked at it
13:25:50linuxstbYes, I want to use the sim to try and trace the common code paths. At the moment I don't have a clue what it's doing.
13:27:31preglowalso, defining out the profiles i don't want in common.h doesn't do anything for the plugin size
13:27:38 Quit amiconn (Read error: 110 (Connection timed out))
13:27:39 Nick amiconn_ is now known as amiconn (
13:27:41linuxstbYou should be able to use DEBUGF - I've defined a macro for it in common.h. We should probably move that to codeclib.h
13:27:49linuxstb(in the sim I mean)
13:27:57linuxstbOr LOGF for the target.
13:28:05linuxstbInside libfaad.
13:29:15 Part LinusN
13:29:21preglowbut i wont have time for rockbox today
13:29:37preglowbut please do tell if you get some analyzing done for libfaad
13:30:02preglowwe most certainly don't need SSR, LTP and LD
13:30:08linuxstbNot sure when I'll get time, but I do want to try and understand libfaad a little better. Maybe tonight.
13:30:08preglowwe MIGHT not need main either
13:30:52linuxstbI'll also install gcc4 for m68k and see if that makes any difference with the ICEs
13:31:00preglowthat would be great
13:31:07preglowi tried compiling gcc4 once, failed miserably
13:31:14preglowbut that was on this box, which is amd64
13:32:20linuxstbI think there is also a problem with the Makefile dependencies for libfaad - so be careful. If you change libfaad, then aac.codec isn't always updated.
13:32:46preglowthink i'll just do some deleting and try again, then
13:33:21preglownot very likely, though, i think the time stamp changed
13:34:14 Join Webguest82 [0] (n=zeroirc@
13:35:32preglowoh well
13:35:40preglowthe gaps are lot smaller than the audio parts now
13:35:43preglowbut still not realtime
13:35:48preglowi expect it'll be a bitch making this realtime as well
13:35:53preglowbut i've got to go for a sec
13:36:57Webguest82Established IRC at to have a Maneunsigan conversation little more with you danger and injury. :)
13:37:31Webguest82-> Established IRC at to have a conversation during little more much times with you danger and injury.
13:38:45B4gderlinuxstb: you prolly need to add a .elf line for the faad codec like the other codecs are listed
13:39:09B4gderhm, no sorry
13:39:24CtcpIgnored 5 channel CTCP requests in 5 minutes and 19 seconds at the last flood
13:39:24*B4gder looks again
13:39:27Webguest82Was H100's firmware made already all?
13:39:29B4gderyes you do
13:39:53Webguest82Was Hangul translation published?
13:39:58*B4gder is confused and will shutup
13:41:14Webguest82Is thing which is made in Korean?
13:42:27 Quit Maxime (Read error: 110 (Connection timed out))
13:48:53 Quit Webguest82 ("Http://www.ZeroIRC.NET ΆΖ Zero IRC ΆΖ Ver 2.8")
14:03:34 Join muesli_- [0] (
14:13:54 Quit ashridah ("Leaving")
14:18:37 Quit muesli_- (Read error: 104 (Connection reset by peer))
14:39:12SlasherMoos: fixed
14:50:39linuxstbSlasher: That commit fixes a seeking problem I had as well.
14:53:20preglowwhat kind of seeking problem?
14:55:08linuxstbIf I was randomly seeking backwards and forwards in the same file, then eventually it would stop playback and the hard disk would stay active. I then had to stop playback and restart it.
14:55:34linuxstbBut that wasn't normal use - I was trying as hard as I could to break it.
14:55:43linuxstbBut now I can't.
15:04:52Slasherlinuxstb: that's good :)
15:19:37 Join tucoz [0] (
15:19:52tucozHi, Slasheri, are you here?
15:21:12tucozI noticed something that is reproduceable(?). When I play a song from an album and press the joystick left during the last seconds of the song to play the song again, the wps do not update.
15:21:26***Saving seen data "./dancer.seen"
15:21:48tucozthis is iriver
15:27:02tucozThis is for instance song 1 out of 10 in an album. I listen to the first song, and want to hear it again and press left during the last few seconds. Then wps is no more updated and the next song starts playing instead of the same track playing again.
15:29:54linuxstbThe problem is that Rockbox has already started decoding the next track. So you are probably trying to seek in the next track, not the one that's being displayed.
15:30:20linuxstbSo I'm not sure what the solution could be. It will be worse if you've got crossfade enabled - if you seek during the crossfade, what is Rockbox supposed to do?
15:30:40tucozlinuxstb: that is probably true. I just noticed that the wps says track 2:14. But, it is not updated.
15:31:01linuxstbI'm not trying to say Rockbox is correct. Just that it's a tricky problem.
15:31:28tucozIt is not that much of an issue, but the wps should not 'freeze' when this happens.
15:32:34linuxstbIf you ignore the WPS problem, what happens? I'm guessing the seek is interpreted as being a seek in the next track.
15:32:45tucozlinuxstb: correct
15:33:36tucozI guess the next track is already in the pcm buffer, and that is why this happens.
15:34:25linuxstbIt's the fact that the codec is in the process of decoding the next track. A seek is processed by the codec - so after the previous track has been completely decoded, you can't seek in it any more.
15:35:57tucozOk, that makes sense. Still, there is a bug with the wps that makes it not updating.
15:39:43linuxstbIt's possible that the WPS information for the previous track has been thrown away by that point. But I don't really know that code very well, so I'm only guessing.
15:40:15linuxstbDoes the WPS display correctly for the next track? Or is that the problem?
15:40:26preglowwps is a bit glitchy
15:40:30preglowespecially for short tracks
15:41:48tucozthe problem is that the WPS is not updating. The information for the previous track is showing. The time shows for instance 3.15/3.18 (3.15 being the time for the press left to happen, and 3.18 the total track length)
15:42:13tucozand track number showing track 2:14
15:43:06tucozwhich is correct. But, nothing else is working. That is no progress bar updates, peak meters, id3 info etc.
15:44:07tucozThe id3 tags showing are those for the track that _were_ playing.
15:56:04 Join lostlogic [0] (
15:59:47 Part tucoz
16:01:47 Quit Moos ("Glory to Rockbox")
16:02:02 Join Moos [0] (
16:59:33 Join DMJC [0] (
16:59:37DMJCum... ok
16:59:44DMJCcan someone help me with warranty info?
16:59:46 Quit DMJC (Client Quit)
17:01:00 Join DMJC [0] (
17:01:09DMJCanyone here from australia?
17:01:14DMJCI need to know some info fast
17:01:22DMJCI need to do a warranty claim
17:01:41DMJCto uninstall rockbox... do I just flash it with the original firmware?
17:01:50DMJCI sat on my iriver, the screen is busted
17:02:03DMJCthe actual unit itself seems fine except the lcd screen
17:02:12DMJCthe glass outer cover of the screen is intact
17:02:49DMJCI also need to know which version of the firmware to use
17:04:04 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
17:04:12crwlthat shouldn't matter
17:04:40crwlflash it with the original firmware and remove the .rockbox directory at root
17:06:45DMJCluckily it's under warranty
17:12:30crwlof course, warranties usually don't cover user-induced damage
17:12:38crwlbut i hear iriver hasn't been too strict on that
17:13:59DMJChow the hell do I flash this thing?
17:14:15DMJCI can't see the option, screen damage makes it unreadable
17:14:16crwli think it's doable with the remote
17:19:26 Join DMJC-L [0] (
17:20:53DMJC-Lok I reflashed
17:21:04DMJC-Land now it doesn't boot unless I hit the reset button
17:21:30***Saving seen data "./dancer.seen"
17:22:34 Join dpassen1 [0] (n=dpassen1@resnet-233-61.resnet.UMBC.EDU)
17:24:15 Join rubberglove [0] (
17:25:20rubberglovehello all (or any!) i was just trying to compile todays build and got this error: make: *** libm4a: No such file or directory. Stop.
17:25:53B4gderdid you re-run configure?
17:26:12rubberglovei'll try it again.
17:27:33B4gderdid you use 'cvs up -d' (the -d being the key) or how did you get the source?
17:28:12rubberglovethe daily builds page.. with the 'latest' source
17:28:26B4gderok, I'll check that...
17:29:18rubbergloveyup. same error. i ran make clean and then configure
17:30:31B4gderright, libm4a is missing
17:31:36B4gderfixed in CVS just now
17:31:49linuxstbThanks - sorry about that.
17:32:07rubberglovecool. thanks.
17:33:18linuxstbIs anyone worried about libFLAC still being in CVS? I was going to delete it, but then thought we might want to try the encoder one day.
17:33:31B4gdernah, leave it around
17:34:54linuxstbThe "apps" directory is almost 40MB now...
17:35:33B4gderTotals grouped by language (dominant language first):
17:35:33B4gderansic: 235877 (95.26%)
17:35:40B4gderperl: 5621 (2.27%)
17:35:40B4gdersh: 3457 (1.40%)
17:35:40B4gderasm: 2513 (1.01%)
17:35:43preglownot bad
17:35:45B4gder(lines of code)
17:36:13B4gderTotal Estimated Cost to Develop = $ 8,812,593
17:36:45B4gdercount and estimate by 'sloccount'
17:36:55 Quit Moos (Read error: 104 (Connection reset by peer))
17:37:03 Join Moos [0] (
17:37:15preglowthat's entire rockbox?
17:37:33preglowcodecs amount to quite a fraction
17:37:53linuxstbI would guess the majority.
17:37:58B4gdercodecs are ~100K lines
17:38:37B4gdertime to go
17:38:39 Quit B4gder ("time to say moo")
17:39:23linuxstbMore useless stats - rockbox.iriver: 224356 bytes, aac.codec: 471428 bytes
17:39:35 Quit DMJC (Read error: 110 (Connection timed out))
17:40:23linuxstbEven removing the bss, I think aac.codec is bigger than rockbox...
17:41:44preglowi wonder how that's possible...
17:41:55preglowthis is the same codec they're stuffing in mobile phones
17:41:59preglowhardware based, sure, but still
17:42:40linuxstbI'm sure we can strip it down a lot - it will just take some time for us to understand it.
17:42:54preglowit's quite a complicated piece of software
17:43:06 Quit DMJC-L (Read error: 110 (Connection timed out))
17:44:48 Join ender` [0] (i=ychat@
17:47:25 Join DMJC [0] (
17:47:37 Nick DMJC is now known as DMJC-sleep (
17:48:01amiconnlinuxstb: I checked the apps/ dir size, including subdirs. 9.97 MB, occupying 12 MB of disk space. Nowhere near 40 MB...
17:48:16 Quit ender` (Read error: 104 (Connection reset by peer))
17:49:17amiconnAll source dirs together: 15.5 MB
17:50:11linuxstbMust be my cluster size then. "du -m" says 39MB
17:51:17linuxstbOops. No - it's the test WAV file I have in there....
17:51:48linuxstbNow it's about 13MB of space.
18:04:46 Join ender` [0] (i=ychat@
18:10:42 Quit rubberglove ("CGI:IRC")
18:32:12 Join DrMoos [0] (
18:32:23 Quit Moos (Read error: 104 (Connection reset by peer))
18:32:29 Nick DrMoos is now known as Moos (
19:13:47 Quit Moos ("Glory to Rockbox")
19:14:56 Join Moos [0] (
19:21:31***Saving seen data "./dancer.seen"
19:30:30preglowlinuxstb: we agree we're not going to support more than two channels unless multichannel is the norm for the codec in question (like ac3), yes?
19:34:13 Join arkascha [0] (
19:35:56 Quit arkascha (Remote closed the connection)
19:43:55preglowamiconn: i've solved the mystery of the slow layer1, it seems
19:46:08preglowamiconn: seems it was a cache issue, it works without boosting for 256kbps layer 1 if i just inline the per sample function that's used in the frame decoding
20:48:43 Join paugh [0] (n=kickback@2001:5c0:8fff:ffff:8000:0:3e03:6822)
20:49:10 Quit ghode|afk (Read error: 104 (Connection reset by peer))
20:49:41 Quit Mxm`Pas`Bien (Read error: 104 (Connection reset by peer))
20:52:19 Join Maxime [0] (
20:55:54 Join ghode|afk [0] (
21:19:07 Quit ghode|afk (Read error: 104 (Connection reset by peer))
21:19:49amiconnpreglow: Nice... Did you check whether this also helps layer 2 performance?
21:20:13preglowamiconn: it can't, layer 1 and 2 decoding is separate
21:20:22preglowamiconn: but if you mean just inlining, i'm going to try that afterwards
21:20:23amiconnOh ok
21:20:51preglowlayer 2 decoding has a similar routine, but it eats three samples and not just one
21:21:04preglowi'll check it out anyway, there's no reason for layer2 to actually be slower than layer3
21:21:13preglowi hate these one way associative caches
21:21:32***Saving seen data "./dancer.seen"
21:22:14amiconnIt's not necessarily a cache issue
21:22:32amiconnYou say this function is called for every sample, i.e. really often
21:22:47amiconnParameter passing is done via the stack on coldfire
21:23:49preglowyes, sure
21:23:54preglowbut i can't imagine it should hurt _this_ much
21:23:55preglowi mean
21:24:06preglowit's actually heavier than a full imdct per band
21:24:10preglowand that's just plain impossible
21:24:28amiconnInlining avoids the whole call overhead
21:24:43preglowit does
21:24:46preglowand it can aid optimising
21:25:04preglowperhaps i should just try compiling the entirety of libmad O3
21:25:16preglowthat should auto-inline functions if it thinks it'll benefit
21:25:27amiconnI can't build the h1x0 win32 sim, I get a really strange error
21:26:02amiconnCC bits.c
21:26:02amiconnIn file included from common.h:261,
21:26:02amiconn from bits.c:28:
21:26:02DBUGEnqueued KICK amiconn
21:26:02amiconnfixed.h: In function `MUL_R':
21:26:02amiconnfixed.h:67: error: `_asm' undeclared (first use in this function)
21:26:02***Alert Mode level 1
21:26:02amiconnfixed.h:67: error: (Each undeclared identifier is reported only once
21:26:04amiconnfixed.h:67: error: for each function it appears in.)
21:26:06amiconnfixed.h:67: error: parse error before '{' token
21:26:08amiconnfixed.h:71: warning: no return statement in function returning non-void
21:26:10amiconnfixed.h: In function `MUL_C':
21:26:12amiconnetc etc
21:26:26amiconnsome more undeclared symbols named "_asm"
21:26:31preglowahh, riiight
21:26:39preglowit probably doesn't check if it uses msvc or gcc
21:26:46preglowlibfaad uses msvc style inline asm
21:26:52amiconnThis is really strange, I can build the x11 sim just fine - also under cygwin, using the very same gcc!
21:27:23preglow#if defined(_WIN32) && !defined(_WIN32_WCE)
21:27:27preglowyes, like i thought
21:27:45amiconnI think that can be safely removed...
21:28:10preglowi think i'm a bit lagged
21:29:20amiconnWay back in time we had an msvc project for building the sim, but that broke long ago, and was removed because nobody fixed it again
21:29:31preglowjust as well
21:29:36preglowcan't see a reason to use msvc
21:33:45preglowapart from the ui, of course
21:33:52preglowthe compiler itself isn't very good
21:36:03***Alert Mode OFF
21:41:39amiconnImho these msvc-style inlines should be removed, and perhaps replaced by gcc-style inline asm for x86
21:41:44amiconnWhat do you think?
21:42:59preglowi agree
21:43:08preglowcan't see a use for the msvc ones, at least
21:43:27preglowthough i'm a bit puzzled by the need for these functions anyway
21:43:44preglowi would have thought x86 compilers were clever enough to actually use those instructions anyway
21:45:16amiconnPerhaps msvc isn't? ;)
21:45:31preglowwouldn't have surprised me
21:45:35preglowit used to be a really poor compiler
21:45:41preglowthough i here the more recent incarnations are pretty good
21:57:20amiconnWhat's so special about these inlines? The use of the imul instruction.
21:57:31amiconn(I'm an x86 asm noob)
21:57:31preglowit does a 32x32->64 bit mul
21:57:43preglowi'm not much of an expert myself
21:57:46preglowbut i know that
21:57:59amiconnIf so, I think the inlines can probably removed
21:58:03preglowand then theres the shrd instruction, that does shifting from one register to another one
21:58:04amiconngcc, uses imul
21:59:19amiconnI disassembled the simulator aac.codec with objdump, and checked a function that uses some MUL_R s
21:59:51preglowdoes it use shrd as well?
22:00:14amiconnThere are seveal imul instruction, and also some shrd
22:00:52preglowi think they can be removed
22:01:04preglowi refuse to believe modern compilers can't do 64 bit muls on x86
22:01:21amiconnx86 asm looks strange
22:01:29preglowand the less i see of the idiot inline assembler thing msvc uses, the better
22:01:31preglowin what way?
22:02:16amiconnReally complex instruction format, and looong instructions
22:02:34amiconnIt seems instructions can be any length from 1 to 10 bytes
22:03:05preglowthe instruction length varies wildly
22:03:36preglowi'm not really fond of x86 assembler myself
22:04:04preglowit got a bit easier with x86-64, since you at least get a semi-decent number of registers to play around with
22:11:15preglowi'll try inlining the layer 2 routine now, and see if it helps
22:11:22amiconnOkay, completely removed these inlines locally. Now the win32 sim builds again :)
22:11:49amiconnShall I commit?
22:14:55preglowamiconn: performance actually seemed to get worse if i inline II_samples
22:16:13preglowbut it's hard to say, at least it didn't get any better that i could see
22:18:36amiconnaac.codec is a monster on the sim as well. 250KB for win32, and 662KB (!) for x11
22:18:55preglowwhat's the size difference?
22:19:10amiconnI'm not sure... perhaps debug symbols
22:19:23amiconnAll codecs are larger for x11 than for win32
22:19:49preglowit doesn't make sense to include debug symbols in a binary image
22:20:10amiconnThe sim codecs aren't plain binary images
22:20:13preglowand the size for windows doesn't make sense either, that's actually almost twice as small as for the actual target version
22:20:17amiconnThey are shared libraries
22:21:42amiconnSome codecs are half the size of the target version in windows, but some are roughly equal in size
22:21:42preglowi'd like for us to use something like that on target as well, one day
22:21:50preglowat least just a simple format with relocation information
22:22:00amiconnThis doesn't correlate to the absolute size, must be something else
22:22:33preglowi think i'll just commit the few libfaad macros i have so far
22:22:51preglowwithout commiting the one that provokes the ICE
22:23:30amiconnYou could put that in as well, and disable it for now (#if 0)
22:28:55 Quit Kohlrabi (Nick collision from services.)
22:28:58 Join Kohlrabi [0] (
22:33:42preglowMUL_R requires me to do a 64 bit shift by 17
22:33:53preglowthink i'll just modify your libmusepack routine a bit for that one
22:34:39 Join ghode|afk [0] (
22:38:32amiconnshift by 17?
22:38:50amiconnShouldn't that be 14?
22:40:14amiconnIiuc this is exactly the same as mpc_multiply()
22:43:40 Join muesli_- [0] (
22:45:22preglowshould be 14, yes, i misread
22:45:37preglowmpc_multiply shifts by 16
22:46:18preglowforget me
22:46:28preglowyes, i can just use the mpc_multiply verbatim, it seems
22:53:46 Join Midgey34 [0] (
23:10:52CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
23:10:52*preglow hopes for a new ICE
23:21:33***Saving seen data "./dancer.seen"
23:48:19 Join RotAtoR [0] (
23:54:04 Quit _FireFly_ ("Leaving")
23:56:54 Quit Kohlrabi (Read error: 104 (Connection reset by peer))
23:58:20linuxstbpreglow: I've just updated to your fixed.h changes, and I am just getting noise out of the decoder now.

Previous day | Next day