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 2007-10-23

00:00:11n1sso does that mac parallell loading work like a regular move.l ?
00:00:15 Quit bertrik ("sleep")
00:01:47preglown1s: yeap
00:01:49preglowpretty much
00:01:58preglowjhMikeS: no new comments on msac and its timing?
00:02:05preglowif not i intend to use it
00:02:07n1sand it's basically free?
00:02:13preglown1s: one cycle extra if from iram
00:02:36preglowmac.l in coldfire is the closest to true dsp instructions you'll come with rockbox
00:02:48n1sand a move.l costs what? 2
00:02:59preglowit depends on operands, anything from 1 to 5
00:03:26preglowthat was wrong, btw
00:03:29preglow1 to 4
00:03:57*rasher wonders why the fireworks demo doesn't start in auto-fire mode by default, instead of showing the menu first
00:04:10preglowrasher: feel free to fix
00:04:56amiconnn1s: move.l from memory costs 2 cycles from iram with all "sane" addressing modes
00:04:56rasherI may just
00:05:07amiconnRegister indexed costs one cycle extra
00:05:29amiconnmove.w and move.b from iram cost 3 cycles though, also with one cycle extra for register indexed
00:05:56amiconnMCF5249UM.pdf, pages 64ff
00:06:34amiconnpreglow: I'm almost sure msac timing is identical to mac. Everything else just doesn't make sense
00:06:41preglowamiconn: i agree
00:07:50preglowooh, finally
00:07:55preglowqmf_synth() is perfect for emac
00:08:02preglowit's even bloody unrolled by four by default
00:09:10n1samiconn: thanks
00:09:14 Quit davina ("xchat on Ubuntu 7.04")
00:09:19preglowamiconn: can you do any faster clips to -32767..32767 with some fancy stuff than i do in that .S?
00:09:24 Quit Toxicity999 ("Leaving")
00:09:26preglowi'll be needing it again, i see
00:09:44amiconnYeah, I already wondered about that ugly clip.
00:10:01preglowi suck at stuff like that :/
00:10:01amiconn10 cycles without clip, 6 cycles with positive clip and 9 cycles with negative clip
00:10:20 Quit n1s ()
00:10:43 Nick fxb is now known as fxb__ (
00:10:44preglowmy only comfort is that gcc sucked even worse :)
00:10:45amiconnUnsigned byte clipping is easier to optimise as you can see in the mpeg2 idct
00:11:01preglowyeah, i know, already had a look at that, and saw no way to utilize that here
00:11:10preglowthat trick only works for eight bit values too
00:11:44amiconnIt also works for 16 or 32 with one extra instruction
00:11:53amiconnBut it works for unsigned
00:12:20preglowwell, if i bias the value by 32767, it can be unsigned, temporarily
00:12:23amiconnThe extra instruction would only be executed in the clipping case
00:12:40preglowthe value of 32767 is also already in a register
00:12:57amiconnFor unsigned we would need 65535
00:13:13preglowi'm incredibly tight on registers at that point too
00:13:20preglowi basically have none free apart from d0 and d1
00:13:33preglowand d0 i can't modify
00:13:36amiconnBut cmp.l #IMM, reg is single cycle as cmp.l reg,reg
00:14:08preglowwhen it's already in cache, yeah
00:14:16preglowand this is a loop, so...
00:14:17amiconnIt's just slightly slower if the instruction pipeline starves because of too many consecutive instructions with extension words
00:14:24preglowahh, right
00:14:49amiconnBut an occasional instruction with extension words doesn't hurt
00:15:13 Quit ender` (" Connection Reset by Gypsies with Wire Cutters")
00:15:23 Quit Rondom ("Ex-Chat")
00:15:39amiconnFor 16 bit unsigned, you would just put an ext.w %d0 after the spl.b %d0
00:16:24amiconn(before the 1: label of course)
00:17:03preglowof course
00:17:04amiconnThat would be 4 cycles both non-clipping and clipping
00:17:17 Join xreigninflame [0] (i=48927114@gateway/web/cgi-irc/
00:17:21amiconnCould pay off even with pre/post adjustment
00:17:22preglowplus biasing and unbiasing
00:17:22rasheramiconn: mandelbrot seems to have a weird blot of colour to the right of the "main feature" (at about 1,0 I would guess). Is this just an oddity of manelbrot, or a bug?
00:17:24preglowwhich would mean 6
00:18:07amiconnYes, and that equals the shortest path of your version, and is just 60% of the most likely path of yours
00:18:15xreigninflameim using vmplayer and i need a .vmx file. where can i get one and which one do i need to help develope RockBox?
00:18:24preglowamiconn: it will almost never clip
00:18:31amiconnYeah, exactly
00:18:49amiconnIn which case your version needs 1 cycles most of the time
00:18:54amiconn*10 cycles
00:19:01amiconnWhere did that 0 go? :>
00:19:12rasherxreigninflame: did you read ?
00:19:37xreigninflamenope musta missed that thanks rasher
00:19:55 Join linuxstb [0] (
00:20:10 Quit J3TC- (Read error: 110 (Connection timed out))
00:20:37 Quit lee-qid ("aufwiederbyebientotsayonara")
00:20:45amiconnpreglow: If you don't need the 32767 in a register, you can do that pre/post adjustment without an extra value in a register
00:21:10preglowamiconn: i don't need it, no
00:21:11amiconnAH, no, forget that
00:21:24preglowok :)
00:22:12 Join iamben [0] (
00:22:17preglowamiconn: reload file and check .order_8, is that right?
00:23:10amiconnUmm, slightly off
00:23:31amiconnAnd you can only clip to [-32768, 32767] that way
00:23:51amiconn(standard int16_t range)
00:24:23amiconnYou would have to add/sub 32768
00:24:38linuxstbjdGordon, kkurbjun: Regarding commit 15258, plugins can be disabled in tools/configure, there is no need to do it in SOURCES/SUBDIRS
00:25:40preglowamiconn: i wonder if that matters
00:26:02amiconnDoes the C source really clip to [-32767... ??
00:26:53 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
00:27:01preglowamiconn: well, it makes sense in some ways
00:28:02preglowamiconn: however, it won't matter much for us, i'll set up macsr to be saturating mode anyway, so if we get any off by one errors, it'll clip
00:28:38preglowamiconn: so, if i change that #32767 to #32768, it should work?
00:30:21preglownow wait up......
00:30:24preglowdoesn't emac have a 16 bit mode?
00:30:59*preglow checks docs
00:32:47mirakamiconn: have you evaluated if the C version implemented in asm without the emac could be as fast as the raw approach with full matrix mutls ?
00:33:23mirakI remember trying one but it crashed so I gave up. Mike S uses the traduction of the C version for arm
00:33:37mirakarm doesn't have a emac like ?
00:34:16preglownot as usable
00:34:29preglowslower and uses ordinary regs for accumulators
00:35:28mirakgood night
00:35:35 Quit mirak (Remote closed the connection)
00:36:15preglowbeh, the 16 bit stuff is only in frac mode
00:37:43preglowamiconn: speex uses 16 bit ints all over, if any -32768 is ever negated, something will puke quite badly
00:38:08preglowamiconn: i guess that's why the clip limits are as they are
00:40:40 Join Mouser_X [0] (n=someone@
00:40:43 Part Mouser_X
00:40:52 Join Mouser_X [0] (n=someone@
00:44:05amiconn16 bit stuff?
00:44:12 Quit obo ("bye")
00:44:13preglowamiconn: 16 bit rounding
00:44:26amiconnah, hmm
00:44:39preglowjust seems to move the rounding point to the middle of the accumulator, for 16 bit output
00:46:19preglowiir_mem16() is very late in the signal chain, though
00:46:24preglownot much processing should happen
00:46:25*amiconn still didn't come up with a clever way to do the 4-path split for the demac vector math
00:48:05 Quit bluebrother ("leaving")
00:48:09 Join qweru [0] (
00:52:44preglowcan't even build a profile build
00:52:51preglowinternal compiler error
00:53:18 Join DogBoy [0] (
00:53:31 Quit Nico_P (Remote closed the connection)
00:58:37 Part linuxstb
01:01:56 Quit xreigninflame ("CGI:IRC (EOF)")
01:03:37 Join Toxicity999 [0] (n=bryan@unaffiliated/Toxicity999)
01:06:09 Join TotallyInfected_ [0] (
01:06:44 Quit TotallyInfected (Read error: 104 (Connection reset by peer))
01:09:11 Join nnk [0] (i=c291b74b@gateway/web/cgi-irc/
01:09:21nnkhi there
01:10:20rasherOdd, idle time is reset when the backlight powers off, so Backlight: 90 seconds, idle poweroff: 1m will see the unit power off after 150 seconds
01:10:32nnkwas wondering what the power management status and the sound quality on the e200 sansa series is curently
01:10:40nnkanybody has any insight on that
01:11:37rashernnk: It still uses more battery than it given better knowledge, but it gets between 10-16 hours battery life depending on file formats, usage patterns etc
01:11:56 Quit Thundercloud (Remote closed the connection)
01:12:06preglowthe speex license seems to require that the copyright statement be included in docs or binary
01:12:11preglowis that gpl compatible?
01:12:29nnkrasher: that sounds very good, i am sure it will get better in a short while, at this rate
01:12:38Zagorpreglow: sounds like standard bsd
01:12:49preglowZagor: okiedoke
01:12:54 Join criznach [0] (
01:12:55rasherpreglow: gpl has some sort of "include copyright statements in binaries under some conditions", so it's probably fine
01:13:03preglowZagor: i'm producing a new file, i'll just use whatever speex uses for a license
01:13:12preglowZagor: in case jmvalin wants to use the code
01:13:25nnkcan i assume both flac and ogg are supported without problems allready on the e200 series? (i seem to understand they are software implementations)
01:14:05rashernnk: Yeah, that works just dandy
01:14:05Mouser_XUsually, once Rockbox gets running on a target, the audio formats also run without a hitch.
01:14:23Mouser_X(Though, it partially depends on the CPU speed of the unit, for some formats.)
01:14:49Mouser_XAs far as I know, the Sansa stuff is running fine and dandy.
01:14:58nnkwell, in that case, i would say the sansa e200 have a groundbraking firmware available, it seems to me it is more than usable
01:15:06***Saving seen data "./dancer.seen"
01:15:07Mouser_XVery true.
01:15:46nnki allready have a player, but i think tomorrow i will have a sansa, allthough it's "not wise". i have a hunch i won;t be able to help myself..
01:16:48nnkthanks to everybody who is working on this in anyway, btw, i was shocked looking through the manual...
01:16:51nnkhats off ;)
01:18:56nnkmouser_x: well, i guess in the case of the e200 series it shouldn;t be a problem, as far as i could tell from the specs, the cpu is quite a beast (200-something mhz, i was running linux on a thinkpad with less than that just one year ago :) )
01:20:52amiconnIt's an 80MHz arm dual core cpu
01:21:06preglowamiconn: did you see anything else stupid in that .S?
01:21:34preglowokiedoke, commiting time, then
01:21:49preglowafter a couple of more tests
01:22:07amiconnI wonder whether we want different optimisations for the different arm targets
01:22:16preglowin what way?
01:23:19amiconnI noticed today that my (not working as I thought, because of alignment) idea how to optimise the vector add/sub would give a large gain on arm7tdmi, but only a small gain on arm920t
01:23:41preglowin which case: yes, i think we should
01:23:46amiconnThis one would probably on par on higher arm versions, but there might be some where it's the opposite
01:24:02preglowon gigabeat s, for example, we also have saturating dsp instructions we want to use
01:24:14 Join hcs [0] (n=agashlin@rockbox/contributor/hcs)
01:25:50amiconnMy idea for arm was using packed add/sub like on coldfire. But arm _demands_ alignment...
01:26:56 Join Soap_ [0] (
01:27:51preglowargh, i want to do qmf_filter(), but i really need to sleep :/
01:27:52 Quit scorche|w ("CGI:IRC (EOF)")
01:28:16nnkamiconn: ahm, okay, i must have confused with something. i tried to find the info again and i couldn't,
01:28:17preglowqmf_synth(), i mean
01:28:25nnka clear sign that i should go to bed ;)
01:28:35rashernnk: sounds like the toshiba gigabeat f
01:29:00amiconnThat would be 300MHz
01:29:02nnkthanks you all a lot, i have a hunch i will be back soon, ;)
01:29:45nnkahm, okay, i found it
01:29:48nnk CPU PortalPlayer 5024 dual core 100 MHz
01:30:24nnkwell, nowhere near 200-something, but anyway, quite close to my single pentium mmx 233mhz (mobile) :))
01:30:58amiconnJust the MHz don't tell much. It's a very different architecture
01:31:03 Quit Zagor ("Client exiting")
01:31:39nnkamiconn: yes, i know, from arm to x86 it;s a long way, i was just making fun of my own statements ;)
01:31:54 Quit jhMikeS (Read error: 104 (Connection reset by peer))
01:32:12 Join J3TC- [0] (
01:32:37 Quit amiconn (" mrf, b0rked sound")
01:33:17 Quit Genre9mp3 ()
01:34:42 Join amiconn [0] (n=jens@rockbox/developer/amiconn)
01:35:29 Join jhMikeS [0] (n=jethead7@rockbox/developer/jhMikeS)
01:37:56preglowamiconn: that clipping stuff actually had a notiable impact :)
01:37:58 Quit Soap (Nick collision from services.)
01:38:03 Nick Soap_ is now known as Soap (n=Soap@rockbox/staff/soap)
01:38:28 Join Soap_ [0] (n=Soap@rockbox/staff/soap)
01:40:37J3TC-rasher, you there? :3
01:40:43J3TC-I get that error still lol
01:41:25rasherJ3TC-: I don't know.. I'm not really motivated to look at it, to be honest
01:45:12J3TC-Lol, ok
01:45:15J3TC-Sudoku ftw tho ;d
01:46:03 Quit petur ("Zzzz")
01:46:30nnkokay, good night
01:46:39nnkrasher: thanks a lot
01:46:49nnksee you guys
01:47:11 Join JdGordon [0] (
01:47:51 Quit nnk ("dd if=/dev/random of=/dev/hda bs=512 count=1")
01:51:46 Quit japc (Read error: 110 (Connection timed out))
01:56:57 Quit Toxicity999 ("Leaving")
01:58:58 Join Toxicity999 [0] (n=bryan@unaffiliated/Toxicity999)
02:00:26 Join TotallyInfected [0] (
02:03:31 Quit TotallyInfected_ (Read error: 104 (Connection reset by peer))
02:05:22 Join rep|icant [0] (n=rep|
02:05:59rep|icanthowdy, any recommendations for music players for automatic syncing back and forth to the rockbox database using a mac?
02:06:15rep|icantin other words, whats the best music player to use with my rockboxed ipod while on a mac?
02:06:34rep|icantpreferably with tag editing,a nd some sort of syncing would be cool
02:07:28krazykitrep|icant, the only thing i know of is amarok, and i think it requires x11 among other things. it is somewhat off topic though
02:07:44rep|icant*nod* ahh, should i take this to rockbox-community?
02:08:03rep|icantok maybe i should look at installing amarok kde/qt
02:11:44 Join Bagder_ [0] (
02:13:27 Part pixelma
02:22:12 Quit hcs ("Leaving.")
02:25:30 Quit Mouser_X (Read error: 104 (Connection reset by peer))
02:25:44J3TC-It compiled
02:26:28J3TC-But I think there's still something wrong lol
02:26:46 Join Mouser_X [0] (n=someone@
02:28:19J3TC-Ok, got it to just 1 warning.
02:28:35J3TC-I'll ask bluebrother for the other one since he was able to fix that warning
02:29:10 Quit Bagder (Read error: 110 (Connection timed out))
02:38:09 Join xreigninflame [0] (i=48927114@gateway/web/cgi-irc/
02:38:16 Join JETC [0] (
02:39:42xreigninflamewhere does the svn store the trunk for vmware player
02:39:54scorchewhere you tell it to
02:40:14xreigninflameumm well what if i cant find it....
02:40:28scorchewhat was teh command you typed to get the source?
02:40:45xreigninflamesvn co svn:// rockbox
02:40:56scorchethen it is in the rockbox folder
02:41:18xreigninflamein root?
02:41:25xreigninflameim in the terminal
02:41:33xreigninflamein the root folder
02:41:36scorchefrom wherever you typed it
02:41:37 Part kugel ("Benutzer ist abwesend.")
02:42:23 Quit donutman25 ("ChatZilla [Firefox]")
02:43:31xreigninflameand i dont have to like add the c language to the vmware do i? when i right click in the enviroment i see a menu called apps and the submenu programming and i dont see c in there
02:44:00krazykitno, everything is set up for you
02:44:28xreigninflameand just to make sure its the xeditor right?
02:44:49krazykitwhat is?
02:44:57xreigninflamewhat i should be editing in
02:45:07xreigninflameim used to using fedora red hat
02:45:13krazykityou can use whatever you want. nano, vim, ed, xeditor
02:45:15xreigninflameso vm is really new to me
02:45:29xreigninflamethanks all
02:46:42 Quit newbyx86 (Nick collision from services.)
02:46:45 Join newby [0] (
02:46:45 Quit xreigninflame ("CGI:IRC (EOF)")
02:50:52 Quit antgel ("Lost terminal")
02:52:28 Quit bb (Nick collision from services.)
02:52:33 Join bb_ [0] (n=bb@unaffiliated/bb)
02:54:07 Quit J3TC- (Read error: 110 (Connection timed out))
02:54:07 Nick JETC is now known as J3TC- (
02:54:53J3TC-bool sudoku_generate(struct sudoku_state_t* state, char* target)
02:55:07J3TC-This one is giving me a warning :3
02:55:21J3TC-Something to do with char* target
03:00:24rashercheck the comments on the patch
03:00:31rasherI'm pretty sure that's mentioned there
03:09:21 Join hannesd [0] (
03:14:48J3TC-Well...that froze up on me
03:15:09***Saving seen data "./dancer.seen"
03:17:03 Join RoC_MasterMind [0] (
03:19:27keanuanyone recall the specs of the Sansa Connect?
03:19:45keanuprimarily SoC info
03:23:27*keanu is just wondering if there isn't a port to the Connect because of hardware, or the fact that nobody is interested
03:26:32J3TC-Hrmm..the sudoku patch works that warning but works.
03:26:42J3TC-A bit slow in generating tho but yeah...better than nothing
03:26:47J3TC-Thanks rasher for the patch lol
03:26:56 Quit Mouser_X (Read error: 110 (Connection timed out))
03:27:15rasherJ3TC-: please add your fixed patch to the tracker entry
03:27:37J3TC-Don't know how :3
03:27:48*J3TC- reads up on making a patch
03:27:58rasherDid you check out using svn?
03:28:28J3TC-Newp. I'm new to this :3
03:28:52J3TC-Although, I think it's better to wait and ask bluebrother for help since I saw his comment on your patch and he was able to fix that specific warning
03:29:13J3TC-He didn't u/l that tho so yeah
03:29:47 Join sarixe [0] (
03:30:20rasherJ3TC-: it's not so much that warning as the menu changes
03:31:20J3TC-Ok, then I'll read up on how to make a patch and all that and get it out sometime tonight. I'd still like to work on that warning but yeah...I guess if it works, it works
03:31:44rasherIf you used svn, do "svn diff"
03:32:19J3TC-Ah, cool
03:32:20J3TC-There it is
03:33:41 Quit courtc (Read error: 113 (No route to host))
03:35:34 Quit Toxicity999 (Read error: 113 (No route to host))
03:36:17sarixequestion: how good is support for ipod video 80gb? and is classic support coming soon?
03:36:41 Join Lars_G [0] (n=Lars@unaffiliated/lars-g/x-000001)
03:36:42Lars_GHi all
03:36:45Lars_Gis svn broken right now?
03:37:17sarixedoesn't look like it
03:37:29J3TC-Good enough?
03:37:34Lars_Godd. I pulled
03:37:45Lars_Gand just in case I removed my nano dir, and reran ../tools/configure
03:38:03sarixejust got: Checked out revision 15274.
03:38:19Lars_GI run make, and get LOTS of errors for: drivers/button.c:
03:38:48Lars_GHuh wait. I had the double click patch in button me thinks. I pray that didn't break stuff
03:39:26rasherJ3TC-: do you hae an account on Flyspray?
03:40:13Lars_GSigh if it's that I'll have to wait. I don't wanna lose that patch
03:40:18Lars_Gand that means adapting it to the new code
03:40:38J3TC-Nope. I'm new to this whole rockbox building stuff.
03:41:45J3TC-You can up it yourself. It's still your code/patch lol
03:41:59Lars_GO.o wtf I got diff comments IN the source files
03:42:04Lars_Gstuff to manually merge perhaps?
03:44:31Lars_Gsigh yeah, it cionflicted
03:47:33Lars_GPhew long live kdiff34
03:48:15 Quit TotallyInfected ()
03:48:41sarixeanyone know how well the ipod video 80gb is supported?
03:48:55 Quit XavierGr ()
03:49:06scorcheit works just fine
03:49:18sarixebecause my brother has a 60gb one
03:49:25sarixeand i installed rockbox on it for him
03:49:40sarixeexcept that it went away somehow, i don't know, something to do with the charger
03:49:50sarixeand the default firmware replaced itself
03:50:25Soapthe default firmware rarely does a full power-off on the firmware.
03:50:47sarixethat's what caused it?
03:51:28SoapSo unless he happened to install a new upgrade of his iPod's Apple firmware, what I suspect has happened is he booted once in to Apple firmware, and it has continued to sleep and awake, never fully powering off, and thus never invoking the bootloader, and thus never rebooting rockbox.
03:51:42sarixeno, he wouldn't have done that
03:52:43sarixe...and is rockbox compatible with the ipod database?
03:52:50SoapHe could very well have done that unintentionally.
03:53:12sarixebut usually itunes asks to update, and usually he doesn't update
03:53:18SoapI find my two possibilities both more probable than Rockbox removing itself.
03:54:25SoapRegardless, Rockbox is not able to read the iTunes database - this does not prevent you from using iTunes, though, as Rockbox is able to build its own database, and thus index all the wacky filenames iTunes uses when placing media on an iPod.
03:54:59 Join Toxicity999 [0] (n=bryan@unaffiliated/Toxicity999)
03:56:05sarixeso... ipod tags all the files then?
03:56:20sarixeand then rockbox can read those tags
03:56:29SoapNo, it will build its database off of the filename if proper tags are not in the files.
03:56:37sarixeman, either way, i'd be using it for ogg, so it doesn't really matter
03:56:45Soap"it" being iTunes in this case.
03:56:55sarixeah, true
03:59:37sarixethe whole reason i'm asking is that i'm considering purchasing an ipod
03:59:43sarixeand i definitely want rockbox on it
03:59:56Soapthe classic is not, and might never be, supported by Rockbox.
04:00:09sarixemight never? why's that?
04:00:14Soapyou would have to buy used, or old-in-stock 5.5th gen.
04:01:12sarixewell i suppose i could _deal_ with ipod's software, and mp3's
04:01:18 Join bryan_ [0] (
04:01:21SoapWhy? Because it is undocumented hardware with totally encrypted firmware. It will be a challenge to run any code on the hardware at all, and then you will need to reverse-engineer the hardware.
04:01:38sarixeand they're not planning on releasing the specs?
04:02:02Galoisapple's gotten much worse with time
04:02:05Soapdare I say it? Hell no.
04:02:08 Quit Toxicity999 (Nick collision from services.)
04:02:12 Nick bryan_ is now known as Toxicity999 (
04:02:21Galoisnot only are they not releasing specs, they're actively obstructing third party software
04:03:07Galoiswith the new ipods, you couldn't even _sync music_ with third party software because apple added a new secret hash
04:03:15Galoisthey had to reverse engineer the secret hash
04:04:16 Join Mouser_X [0] (n=someone@
04:04:16sarixeand that 160gb classic looked like SUCH a good deal...
04:05:08sarixethe question is... why are they so hostile to open development?
04:05:54GaloisI really have no idea. Their current strategy is a money loser for sure.
04:06:15rasherJobs likes control
04:06:16GaloisI mean, let's put it this way. People have bought apple products in the past on my recommendation.
04:06:23Soapoh yea, Apple is losing tons of money. WTF!
04:06:30Galoisabout a dozen laptops at least, among my friends and family
04:06:39Galoisthey won't get that money any more
04:06:46Galoisnot with the way they're heading now
04:06:48sarixewell at least they don't have secret filesystems... or do they?
04:07:02sarixeare hfs+ specs released?
04:07:08sarixeor is that even what they use?
04:07:16sarixei'm so out of the loop when it comes to apple tech
04:07:20SoapRegardless, let us keep this channel on-topic, and move all economic discussion to #rockbox-community
04:08:41 Join JETC- [0] (
04:13:13 Quit Lars_G ("Bye bitches :D")
04:13:47 Join TotallyInfected [0] (
04:16:58 Join eigma [0] (n=cat@
04:27:06 Quit miepchen^schlaf (Read error: 110 (Connection timed out))
04:27:17 Join miepchen^schlaf [0] (
04:29:03JdGordonhey eigma
04:31:23JdGordondid you try runnig rockbox with the jtag?
04:32:45eigmano, sorry
04:33:01eigmakarl was going to use my jtag rig remotely, but he had to leave
04:33:04 Quit Rob2222 (Read error: 104 (Connection reset by peer))
04:33:26JdGordonah ok
04:34:56 Quit sarixe ("(EE) Failed to load "quit" module")
04:35:05JdGordonany luck with the backlight?
04:35:19eigmano more than I did before
04:35:26JdGordonoh well :)
04:35:29eigmaI've mapped out a lot of the OF functions that control it
04:35:33 Join TheDarkOne[lappy [0] (n=marc@
04:35:41eigmaso in a few hours I could probably jot down a program that does the same thing
04:35:44JdGordonI got the battery reading working so im going to run a battery bench to see how the values change
04:36:55TheDarkOne[lappyi charged my ipod and it booted into an apple screen with the backlight on, and i can't get out of it into anything else, apple or rockbox
04:37:07TheDarkOne[lappyany ideas how to fix? i tried the reset, did nothing
04:37:10eigmabut I still don't know what chip it is..
04:38:29TheDarkOne[lappyif anyone is hee
04:40:48 Join XavierGr [0] (
04:41:55eigmaTheDarkOne: looks like you're out of luck for now :\
04:42:57 Quit qweru ("moo")
04:48:44 Join psycho_maniac [0] (i=psycho_m@
04:58:14 Quit webmind (Read error: 110 (Connection timed out))
04:59:07 Quit RoC_MasterMind ("Leaving")
05:04:02 Join rep|icant_ [0] (n=rep|
05:05:20 Join rep|icant__ [0] (n=rep|
05:09:53 Quit rep|icant (Read error: 110 (Connection timed out))
05:15:12***Saving seen data "./dancer.seen"
05:15:58 Join bb [0] (n=bb@unaffiliated/bb)
05:19:36 Join hannesd_ [0] (
05:20:16 Quit rep|icant_ (Success)
05:23:42 Quit hannesd (Read error: 110 (Connection timed out))
05:23:42 Nick hannesd_ is now known as hannesd (
05:27:41 Quit bb_ (Read error: 110 (Connection timed out))
05:38:03 Quit XavierGr (Read error: 110 (Connection timed out))
06:11:32 Join BHSPitMonkey [0] (n=stephen@unaffiliated/bhspitmonkey)
06:12:40 Quit jhMikeS (Nick collision from services.)
06:12:46 Join jhMikeS [0] (n=jethead7@rockbox/developer/jhMikeS)
06:15:25 Quit jhulst (Remote closed the connection)
06:15:33 Quit grndslm ("Leaving")
06:23:10 Quit lazka (Remote closed the connection)
06:24:01 Join emeraldd [0] (
06:24:40 Part emeraldd
06:28:22 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
06:29:24 Join Rob2222 [0] (
06:33:57 Quit Calcipher ("I-n-v-i-s-i-o-n 2.0 Build 3515 with A Pack Fix By")
06:43:04 Quit eigma ()
06:56:06 Join markun_ [0] (n=markun@rockbox/developer/markun)
06:56:35 Quit markun (Read error: 113 (No route to host))
07:15:13***Saving seen data "./dancer.seen"
07:15:42 Nick fxb__ is now known as fxb (
07:39:06 Quit Toki (Nick collision from services.)
07:39:19 Join Toki_ [0] (
07:41:13 Join sounddude [0] (
07:42:47 Quit Isolinear (Read error: 104 (Connection reset by peer))
07:42:52 Quit jhulst (Read error: 113 (No route to host))
07:46:31 Nick fxb is now known as fxb__ (
07:48:31 Join rep|icant [0] (n=rep|
07:51:47 Quit Seed (Nick collision from services.)
07:51:55 Join Seed [0] (
07:52:30 Join rep|icant_ [0] (n=rep|
07:56:11 Quit rep|icant_ (Client Quit)
08:03:06JdGordonwaiting for batt bench to finish is boring:'(
08:04:15 Quit rep|icant__ (Read error: 110 (Connection timed out))
08:05:00 Join kubiix [0] (
08:06:22 Join Buschel [0] (
08:07:49 Quit rep|icant (Read error: 110 (Connection timed out))
08:09:15 Nick markun_ is now known as markun (n=markun@rockbox/developer/markun)
08:20:25 Join ender` [0] (
08:25:13 Join midkay_ [0] (
08:31:22 Nick fxb__ is now known as fxb (
08:32:18 Quit BigBambi (Read error: 113 (No route to host))
08:33:09 Join midkay__ [0] (
08:33:27 Quit midkay (Nick collision from services.)
08:33:29 Nick midkay__ is now known as midkay (
08:36:18 Quit Buschel ()
08:37:00 Quit J3TC- (Read error: 104 (Connection reset by peer))
08:37:04 Quit JETC- (Read error: 104 (Connection reset by peer))
08:37:20 Join JETC- [0] (
08:37:49 Join J3TC- [0] (
08:38:21 Join spiorf [0] (
08:38:36 Join Rob222241 [0] (
08:40:30 Join CaptainSquid [0] (
08:42:33 Quit CaptainSquid (Client Quit)
08:50:15 Quit midkay_ (Read error: 110 (Connection timed out))
08:50:40 Join Thundercloud [0] (
08:56:53 Join pondlife [0] (n=Steve@rockbox/developer/pondlife)
08:56:54 Quit Rob2222 (Read error: 110 (Connection timed out))
09:10:16 Join advcomp2019_ [0] (n=advcomp2@
09:10:41 Join Isolinear [0] (
09:10:45 Join petur [0] (i=d4efd6a6@rockbox/developer/petur)
09:10:49 Quit advcomp2019 (Nick collision from services.)
09:10:51 Nick advcomp2019_ is now known as advcomp2019 (n=advcomp2@
09:11:22*petur has been terribly stupid
09:11:47peturcan somebody remove that ban that preglow did yesterday evening?
09:11:57 Join Zagor [0] (n=bjorn@rockbox/developer/Zagor)
09:12:34peturpondlife: here?
09:13:25GodEater_ban ?
09:14:29peturyes, very stupid: I got kicked everytime and thought my account had been hijacked. Now I discovered I left my irc client at work logged in :(
09:14:52GodEater_I do that all the time
09:15:16***Saving seen data "./dancer.seen"
09:15:18peturI don't but due to stress I forgot to kill it when rushing home
09:15:56GodEater_scorche was around a few minutes ago
09:15:59GodEater_he should be able to fix it
09:16:32peturI feel so stupid that I didn't think about this...
09:16:55*petur kicks forehead multiple times
09:17:21GodEater_chill - it's no biggie!
09:17:38GodEater_poor forehead will be in rasher's stats again
09:17:51 Nick Bagder_ is now known as Bagder (
09:18:05GodEater_or Bagder can mend it :)
09:18:19Mode"#rockbox +o Bagder " by ChanServ (ChanServ@services.)
09:18:50Mode"#rockbox -b *!* " by Bagder (n=daniel@rockbox/developer/bagder)
09:19:11Mode"#rockbox -o Bagder " by Bagder (n=daniel@rockbox/developer/bagder)
09:19:11*petur bows to op
09:20:38 Quit petur ("back to normal irc client")
09:20:53 Join petur [0] (
09:25:08*GodEater_ thinks petur should just link a few more nicks to his primary one and not worry about it in future
09:25:35peturI have 4, but I was worried my account was hacked
09:26:25peturand the reason I thought so was because last week somebody else logged in with my nick several times (but not identified)
09:26:38GodEater_get a weirder nick ;)
09:27:01 Nick petur is now known as SatanEater_ (n=petur@rockbox/developer/petur)
09:27:09 Nick SatanEater_ is now known as petur (
09:27:38GodEater_probably a more popular nick with American Midwestern people than mine is ;)
09:27:53peturI know a better one...
09:27:58 Nick petur is now known as beerlover (
09:28:05 Nick beerlover is now known as petur (
09:28:08psycho_maniacpetur_pumkin_eater, kinda long though.
09:29:34pondlifepetur: I'm back
09:29:53pondlifeAny bootloader news?
09:30:01peturI did a bit of diskswapping yesterday and the result puzzles me
09:30:19peturthe 80GB disk in my h320 boots ok
09:30:32pondlifeWith SVN bootloader?
09:30:36peturso it comes down to the 80GB disk in my h340
09:30:41 Join ddalton [0] (
09:30:46 Join davina [0] (
09:30:52 Part ddalton
09:30:55peturwith the last one XavierGr made me some time ago
09:31:06pondlifeSo, the same disk fails to boot in the H340, but works in H320?
09:31:12peturI left my notes at home :(
09:31:42peturbut a 20GB or 40GB disk in the h340 also boot ok
09:32:13pondlifeHave you done a disk info check in the H340 with both disks?
09:32:16peturmaybe we're on the edge of some timing, or my h340 is out of spec hardware wise
09:32:35pondlifeProbably the former
09:33:36pondlifeI don't see how the newer bootloader would make things worse though
09:33:48pondlifeIIRC, the main change is some extra delays
09:34:02pondlifeDo you get an error in the failure case, or a lock?
09:34:29peturit locks after printing battery value
09:35:51pondlifeAre you happy building your own bootloader?
09:36:04peturnot yet ;)
09:36:11pondlifeAh :)
09:36:30pondlifeOK, I won't suggest use of printf :)
09:36:30peturwell I have a backup unit now so I should just do it
09:37:00pondlifeI expect it's the ata_init()
09:37:11pondlifeBut it would be good to know for sure
09:38:15peturI need more spare time first :(
09:39:39pondlifeMe too, no rush
09:42:22 Quit psycho_maniac (" Asta La Vista")
09:42:36 Quit TotallyInfected ()
09:43:46 Join CaptainSquid [0] (
09:47:37 Quit atsea-22 (Remote closed the connection)
10:03:20 Quit criznach ("ChatZilla [Firefox]")
10:03:46 Join pixelma [0] (i=pixelma@rockbox/staff/pixelma)
10:08:40 Quit Slasheri ("Reconnecting")
10:08:47 Join Slasheri [0] (i=miipekk@rockbox/developer/Slasheri)
10:09:43 Quit Isolinear (Read error: 104 (Connection reset by peer))
10:12:00 Join safetydan_ [0] (n=safetyda@rockbox/developer/safetydan)
10:12:11 Nick fxb is now known as fxb__ (
10:15:05 Quit kubiix (Read error: 104 (Connection reset by peer))
10:15:25 Join kubiixx [0] (
10:15:25GodEater_pondlife: re: follow playlist, is it only broken on the archoses ?
10:17:22pondlifeGodEater_ No,. here on H340 too
10:17:31GodEater_well see my post then...
10:17:35pondlifeAuto-change dir doesn't update it
10:18:07GodEater_cos it works fine for me
10:18:10pondlifeMight be intermittent, I only tested it once and it failed the test
10:18:19pondlifeWouldn't surprise me.
10:18:24GodEater_I use it all the time
10:18:29pondlifeI'm running MoB at the moment, so can't retest
10:18:31GodEater_and it hasn't ever gone wrong for me
10:18:36GodEater_and I'm ALSO using MoB
10:19:09pondlifeDo you have dircache enabled?
10:19:21pondlifeOK, me too
10:19:56GodEater_works fine on my other two targets as well
10:20:05pixelmaGodEater: it works for me while music is playing - but if stopped (and without auto-change directory) not the last song I played is selected but the first in the directory
10:20:27GodEater_doesn't stop currently clear the playlist though ?
10:20:59safetydan_preglow, are you subscribed to the speex dev list?
10:21:52pondlifeAh, just found a nice MoB bug.
10:22:04GodEater_Nico will be pleased
10:22:05pixelmaGodEater: no, not anymore. But I just tried on my Sansa and it seems to work now, will keep an eye on it
10:22:40GodEater_just so we're clear on what we're looking at here
10:22:52GodEater_we're talking about Follow Playlist not working when you have Auto-change directory ON?
10:23:18GodEater_that's how I read the forum post in question anyway
10:23:21 Join lee-qid [0] (
10:23:22pixelmaI don
10:23:26pixelma't use it
10:23:28pondlifeI just reproed the bug with MoB
10:23:42GodEater_well as I say it works fine here =/
10:23:49pondlifeWhen you stop playback.
10:23:55pixelmaok, then it's an auto-change directoryy thing
10:24:04GodEater_and you mean *stop* not pause ?
10:24:30GodEater_last playing track looks to be selected ok here
10:24:51pondlifeOK, play the last track in album A, let it move onto album B. Then press STOP.
10:25:05pondlifeDoes it in MoB too
10:26:08GodEater_ah ok - yes, that leaves it on the last track, not the one that just started
10:26:52*pondlife wishes that we could be clear that it's starting a new playlist that clears an old one, not pressing STOP.
10:27:01pixelmaah no (sorry for taking back the taken back). Still doesn't work correctly without auto-change directory - for me the last track I selected is highlighted not generally the first in the directory which got me confused
10:27:03pondlifeSeems a bit halfway at the moment
10:27:31GodEater_yeah, hitting select to show the current playlist works fine
10:27:40GodEater_also, having DONE that, hitting stop shows the correct file too
10:28:00 Join LinusN [0] (i=linus@rockbox/developer/LinusN)
10:28:01GodEater_so it's some pecularity with stop
10:28:03pondlifeSomething's nearly right
10:28:50pondlifeGodEater_: On MoB, if you enable Auto-change dir and turn repeat off, if you play the last track in a folder through, does it auto-advance?
10:29:00pondlifeMine plays the same track again
10:29:10pixelmaI think JdGordon something why it's not easily possible with the current code... if I just could remember
10:29:13pondlifeBut the WPS is a bit confused
10:29:19GodEater_well the only way to get Autochange to work at all is to have repeat set to off
10:29:23GodEater_otherwise it doesn't do anything
10:29:32pondlifeIndeed, just checking
10:29:54 Quit maddler (Connection timed out)
10:29:58pondlifeI selected track 19 in an album - the following album has 10 tracks.
10:30:25pondlifeI'm now hearing Album A Track 19 again, with the WPS displaying the right title, but as track 1 of 10
10:30:56pondlifeThen, after one repeat, it's moved onto Album B Track 1, displayed as track 2 of 10
10:31:17GodEater_are you crossfading or anything else ?
10:31:30pondlifeCrossfade is enabled but only for skip/random
10:31:35pondlifeSo shouldn't be involved here
10:32:26GodEater_hard lock when I change cross fade settings again
10:36:16GodEater_'tis a crossfade thingy
10:36:34GodEater_I just get the symptoms you describe, but only after having the same crossfade setting as you enabled.
10:36:43pondlifeAs opposed to what?
10:36:51pondlifei.e. What worked for you?
10:36:52GodEater_as opposed to no crossfade at all
10:37:33 Join lee-qid_ [0] (
10:37:33 Quit lee-qid (Connection reset by peer)
10:37:48GodEater_playback absolutely refuses to work after a change to those damn settings those
10:37:59pondlifeYou need to reboot
10:38:10pondlifeChanging any crossfade settings really buggers up MoB
10:38:48GodEater_I assume the reason it's stuffing up the autochange in your case is because you're using track-skip to change to the last track in the folder
10:38:54GodEater_before letting it do the auto-change
10:39:04pondlifeAh, but I'm not
10:39:10 Quit sup (Read error: 110 (Connection timed out))
10:39:11pondlifeI'm starting playback on the last track
10:39:43GodEater_ah ok
10:39:46GodEater_I take it back then
10:41:09GodEater_yep - confirmed. No crossfade at all results in perfect change
10:43:06pondlifeSmells like an </ <= type of thing
10:43:33*GodEater_ dribbles a bit and wonders what that means
10:43:43pondlifeLess than vs. less than or equal
10:44:24GodEater_if you say so :)
10:46:08 Nick lee-qid_ is now known as lee-qid (
10:53:55 Join atsea-22 [0] (i=atsea-@gateway/tor/x-8f14cbc6e80998f5)
10:56:53amiconnGodEater: The problem with 'follow playlist' is most probably that right now the wps sets the current browser position only when actively leaving it when music is playing
10:57:16amiconnBack when it was working properly, it did set the position on every track change.
10:57:25amiconnThat behaviour needs to be restored
10:58:17amiconnRight now 'follow playlist' never works when playback stops at end of list
10:58:30amiconn(both hwcodec and swcodec)
10:58:53pixelmaalso not when stopped by hand
11:00:36GodEater_how else would it stop ?
11:00:52amiconnAt the end of playlist...
11:01:01GodEater_sorry - was thinking of auto-change dir again
11:01:09*amiconn never uses auto-change dir
11:01:21*GodEater_ never not uses it
11:01:28amiconnAnd 'follow playlist' has been broken for months now :(
11:01:59GodEater_I've not noticed, since I never use "Stop"
11:02:07GodEater_in which case, it works just fine
11:03:11amiconnHow do you stop then? :>
11:03:15GodEater_pause :)
11:03:23GodEater_or turn it off
11:03:35amiconnI *always* stop before shutting down
11:03:37pixelmaGodEater: are you going to update your post then or should I comment on this? ;)
11:03:39 Join bistouri [0] (i=c1335ac7@gateway/web/cgi-irc/
11:03:48GodEater_it's impossible to not stop when turning off with the ipod I think
11:04:02GodEater_pixelma: by all means comment ;)
11:04:18GodEater_your posting is more eloquent than mine anyway
11:04:30amiconnOn iriver you can - but that's not the problem.
11:05:42GodEater_indeed - but I never stop playback on my iriver when I power it off
11:05:45GodEater_don't see the point
11:06:23amiconn'Follow playlist' not working annoys me when I start an album in car, arrive at my destination, stop & shutdown when the album isn't finished. When resuming later, and the album then ends, it drops to the root dir...
11:07:42 Join japc [0] (n=japc@
11:09:58 Quit CaptainSquid ("Miranda IM!")
11:10:50 Join CaptainSquid [0] (
11:11:03 Quit bistouri ("CGI:IRC")
11:15:18***Saving seen data "./dancer.seen"
11:28:56 Quit idnar (Nick collision from services.)
11:28:58 Join idnar_ [0] (n=mithrand@unaffiliated/idnar)
11:33:58 Join maddler [0] (
11:39:08 Join barrywardell [0] (
11:45:35 Join lachlan [0] (i=7cab80fe@gateway/web/cgi-irc/
11:45:46 Quit lachlan (Client Quit)
11:46:00 Join lachlan [0] (i=7cab80fe@gateway/web/cgi-irc/
11:47:34 Part pondlife ("Gone")
11:47:48 Join kugel [0] (i=kugel@unaffiliated/kugel)
11:48:00 Quit barrywardell (Remote closed the connection)
11:48:38 Join barrywardell [0] (
11:49:01 Join xazax [0] (
11:49:17 Quit bumper (Read error: 110 (Connection timed out))
11:54:09 Join Exiro [0] (
11:55:19Exirocould somebody help me pleas?
11:56:32pixelmawelcome... this question is impossible to answer because we don't know the problem yet ;)
11:56:34peturwe could if you would ask a question
11:57:51Exirowell, when I connect my 1st gen ipod nano with rockbox current build on it to my comp
11:58:00Exirothe screen comes up first
11:58:08Exiroand then it says "do not disconnect"
11:58:22Exiroshouldn't I be able to use it while connected?
12:00:10xazaxThe ipod? you want it connected and listening to it at the same time?
12:00:20Exirouh yes
12:00:32markunin rockbox?
12:00:35xazaxI dunno if thats possible.
12:00:56markunif yes: hold menu while inserting the cable (it will charge only and you can still use the player)
12:01:27Exirooh, just wondering because the normal "don't disconnect" screen comes right after the rockbox screen when I connect
12:01:46Exirobut I am reinstalling it right now
12:02:18markunreinstalling rockbox?
12:02:23pixelmawhat happens when you put it in disk manually?
12:02:54 Join Rondom [0] (
12:03:00pixelma"disk mode" even
12:05:05Exiroyes, i'm reinstalling rockbox
12:07:43 Quit jhMikeS (Nick collision from services.)
12:07:49 Join jhMikeS [0] (n=jethead7@rockbox/developer/jhMikeS)
12:08:57 Join bluebrother [0] (i=tUAqppnE@rockbox/staff/bluebrother)
12:09:11 Join Nico_P [0] (n=nicolas@rockbox/developer/NicoP)
12:09:52 Nick fxb__ is now known as fxb (
12:12:11 Join A_M [0] (
12:21:11 Join n1s [0] (
12:22:39Exirook I'm listening to my music again, but I hear strange sounds in the background
12:22:58Exirolike someone's scratching
12:23:55 Join mr_pink [0] (
12:23:58JdGordonpixelma: what did i say waiiiiiiiiiiiiiiii wththe current code? /me couldnt follow the convo
12:24:07*JdGordon kills his keyboard
12:24:15JdGordonwas impossible with*
12:25:07Exirohey, I hear sounds like someone is swallowing in the background
12:25:17Exirodoes anyone else experience this?
12:26:29Exirooh, and I can't get out of the maaze!
12:28:32mr_pinkI'm having some problems setting up the build environment. I ran the automatic script and thing's seemed to go ok until it failed to download one part. see: Should I just try running it again?
12:28:53 Part mr_pink ("Ex-Chat")
12:29:02pixelmaJdGordon: something about why follow playlist couldn't work when playback is stopped, must be a while ago but I thought you explained that once (though I could remember wrong)
12:29:04 Join mr_pink [0] (
12:29:37JdGordonah yeah, I was supposed to do something about that :p
12:30:13n1smr_pink: that file (the multilibs patch seems to be missing from the download server...
12:30:13pixelmasee... and all the times you complained about being bored.. :P
12:30:17n1sZagor: ping?
12:31:45mr_pinkn1s: Is that the reason the rest of it failed? Shouldn't the script abort if something vital fails to download?
12:31:50Zagorn1s: pong
12:32:07n1sZagor: gives 404 ,maybe?
12:33:04pixelmaExiro: from your description I think your Nano could suffer from this bug (note that there are differences how strong people are affected)
12:33:05Zagorhmm, how did I miss that?
12:33:11ZagorJdGordon: no it's www
12:35:48pixelmaZagor: twhile you're at it, I also wondered about this report
12:35:49Zagorah, I didn't miss it. it's recently added
12:36:19n1smr_pink: no, i guess something is up with your native gcc as it throws "C compiler cannot create executables."
12:37:05Zagorpatch added now
12:37:54mr_pinkn1s: Hmm, any ideas? I'm on ubuntu
12:38:10n1smr_pink: have you installed a native gcc?
12:39:05 Join pepie34 [0] (
12:39:16barrywardellpixelma, Zagor: I think we should just keep the newer PP5022.mi4
12:40:04BagderI think that's due to the particular mirror not deleting deleted files in the source server
12:40:25ZagorBagder: no, has both files too
12:40:43Bagderthen it was just my mistake I guess
12:41:02Zagoryou were in china when that happened... :)
12:41:28ZagorI'll remove the older file
12:42:24Exirook, noob question but... how do I add music from my comp to my player?
12:42:48mr_pinkn1s: Yep, looks like it
12:42:50n1sExiro: drag and drop in your favourite file mangaer
12:43:45GodEater_or least favourite file manager if you're feeling masochistic
12:44:19 Join TentaclePorn [0] (i=Tentacle@
12:44:25TentaclePornHi there ;D
12:44:39ExiroGodeater: is that something like emo? then no!
12:44:46TentaclePornI was wondering if there were any iriver RockBox users out there that may be able to help me out for a second?
12:45:00n1smr_pink: you need at least the packages "gcc" "build-essential" and "binutils" do you have all of them?
12:45:00GodEater_TentaclePorn: difficult to tell since we don't know what your problem is
12:45:06TentaclePornMy bad
12:45:29*n1s has an iriver
12:45:51 Nick kugel is now known as kugel|afk (i=kugel@unaffiliated/kugel)
12:45:58*GodEater_ still thinks we should +m the channel and only voice people when they can prove they've read the guidelines...
12:46:16xazaxTentaclePorn: shoot... whats your problem
12:46:23xazaxGodEater: :D
12:46:35TentaclePornIn any case, the problem is that I can't get my database to, well, do anything. First time I booted rockbox I went to database and it searched through, found all the files, and went back to the menu. Although now whenever I click on database It just says "building database" and sits on 0 found
12:46:47TentaclePornSorry to run in and immediately ask a question =/
12:47:18n1sTentaclePorn: go into the "tagcache" entry in the debug menu and tell us what it says
12:47:33Exiroum drag to what? which folder?
12:47:42xazaxTentaclePorn: I had issues in the beginning... I seemed like restarts magically sorted that out.
12:47:42 Quit J3TC- (Read error: 113 (No route to host))
12:47:59n1sExiro: any folder
12:48:07 Quit A_M ()
12:48:09mr_pinkn1s: Ah, some of the bits from build essential were missing, ill try again, cheers
12:49:30 Quit lachlan ("CGI:IRC (EOF)")
12:49:38TentaclePornI can't seem to find tagcache
12:49:54TentaclePornI have "view dircache info"?
12:50:02GodEater_no - "view database info"
12:50:07TentaclePornoh okay
12:50:15*GodEater_ wonders how old the build n1s is running is :)
12:51:16n1sGodEater_: so the bastards renamed it! but now... it's... al consistent and no fun :-(
12:51:53TentaclePorninitialised: Yes , DB Ready: No , RAM Cache: No , RAM: 0/0 B , Progress: -1% (0 entries), Curfile: −−- , Commit Step: 0 , Commit Delayed : no.
12:52:33GodEater_that doesn't look very healthy
12:52:41*xazax still votes for restart and wait for automagical solution
12:53:02TentaclePornI tried the restart, multiple times ;]
12:53:03n1sTentaclePorn: delete all *.tcd files (if they are still named that?) in your .rockbox dir and start over
12:53:09 Join pondlife [0] (n=Steve@rockbox/developer/pondlife)
12:53:10*xazax curses
12:53:18TentaclePornJust hit delete, unplug, and reboot?
12:53:56Exirowere are the files that are in the database?
12:54:01n1sand init the databse again with fingers crossed ;-)
12:54:05 Join Isolinear [0] (
12:54:23TentaclePornProbably inside the iriver's natural music folder, let me have a look at it again and see how it looks
12:54:52GodEater_Exiro: the files which the database indexes are the music files on your player. The files which *form* the database are all named *.tcd, and live in the .rockbox directory
12:56:07xazaxlive? scary..
12:56:08Exirobut those are only 1/2 kb!
12:56:27TentaclePornn1s: where will I find all the .tcd files
12:56:39n1sTentaclePorn: inside the .rockbox dir
12:56:47TentaclePornI found one, that seems to be it
12:57:11GodEater_Exiro: how much music do you have on your player ?
12:58:10 Quit lee-qid (Read error: 110 (Connection timed out))
12:58:26Exirolike 100 songs I guess
12:58:53TentaclePornI thought I was a small fry with 2000 D:
12:59:12GodEater_Exiro: that's why they're so small then
12:59:47Exiro? they are normally 2mb
13:00:06Exiroi mean 6
13:00:20GodEater_you think your database should be 6mb in size for only 200 songs ?
13:00:25TentaclePorndatabase is finding files again! Thanks n1s lets hope this works
13:01:09xazaxTentaclePorn: so... Youre databasefile was corrupted?
13:01:12bluebrotherpixelma: sure mpegplayer can seek? AFAIK it can only resume
13:01:25Exiro1 song in mp3 format is 6 mb
13:01:39GodEater_so ?
13:01:46GodEater_the metadata in it is tiny
13:01:50TentaclePornWell, the database build completed, told me to reboot. I did, and now when I click database it still says ""database is not ready, initialise now"
13:01:53Exirowere is that mp3 file on the ipod nano?
13:02:04GodEater_what mp3 file ?
13:02:09Exiroon the ipod
13:02:21GodEater_I have ~8000 tracks and my database isn't even 1MB
13:02:32GodEater_what mp3 file are you talking about ?
13:02:35Exiroif the song on my comp is indexed, it still needs the songs itself on the ipod
13:02:38pixelmabluebrother: you can choose "set start time" which is basically like seeking, just not very convenient
13:03:04Exirowhen it plays a song, it needs an mp3 to read right?
13:03:21bluebrotheroh, nice. Haven't noticed that.
13:03:23GodEater_yes - but you were told earlier to copy it to your ipod
13:03:32TentaclePornHmm, now I have the same problem again, it just sits on "building database - 0 found"
13:03:33GodEater_how the heck should I know where you copied it to ?
13:03:40bluebrotherthough that screen is flickering like hell on my mini
13:03:50Exirono, i didn't copied
13:04:02GodEater_Exiro: then the mp3 isn't there is it ?
13:04:12GodEater_TentaclePorn: that could be due to a damaged music file
13:04:15pixelmabluebrother: yeah, it's not nice and the greyscale lib makes it worse I think
13:04:28Exiroright away, I removed all files from my ipod, then I installed Rockbox to it, the bootloader, fonts
13:04:36n1sTentaclePorn: could you do the same again with the delete bug go into that debug screen while it builds the database and see which is the last file it opens?
13:04:41Exiroand when I started it, it had a database with all my songs
13:04:59Exiroand I didn't copied them, they weren't in 'files'
13:05:13TentaclePornn1s: Sure thing, thanks for the help so far, by the way ;D
13:05:37Exirowere did that database come from?
13:06:17pixelmaprobably from the files you put on your Nano through iTunes
13:06:39Exirook, but where are those files now?
13:07:18Exirooh I get it
13:07:29Exirothe hidden folder "ipod controll"
13:07:37pixelmayou could see them in the file browser too but iTunes stores them in a folder that is hidden by default, you would need to set the "file view" option in Rockbox to "all" and look inside the ipod_control...
13:07:54Exirolol I was first
13:08:01Exirook thx now I get it
13:08:34 Quit pepie34 ("Ex-Chat")
13:08:36ExiroAnd if I replace those mp3's, the database doesn't work anymore right?
13:09:23n1sExiro: you should probably take a loot at the manual
13:09:59TentaclePornn1s: Doing it now, am I meant to be watching the Curfile?
13:10:36TentaclePorncool. It seems to be going considerably slower this time, Im assuming thats because its running in the background
13:12:46jhMikeSkkurbjun: no plugins seem to function on gigabeat now :[
13:12:57 Join Arathis [0] (
13:13:30preglowsafetydan_: no
13:14:27safetydan_preglow, not really important, just someone posting changes to get speex running on the Cell SPEs. They look like they need the same sort of changes we do for allocating decoding structs in special ram.
13:14:46preglowi'll check it out
13:14:49preglowand yeah, they dio
13:15:19***Saving seen data "./dancer.seen"
13:15:57 Quit xazax ("Leaving")
13:16:09jhMikeSkkurbjun: I take that back. Some don't. fire, mpegplayer, blackjack (causes reboot), pacbox. some seem to work but can't say they actually do properly.
13:17:02TentaclePornn1s: I never caught the very last file as the light went off, but it had finished scanning the music and had started going through the .rockbox/* folders
13:17:43n1sTentaclePorn: well, I'm out of ideas then :-(
13:18:02TentaclePornthat's okay, thanks for the help
13:18:41TentaclePornlet me just make sure I have the right idea here - after initialising the database, I should be able to click on database and browse my artists and albums etc., am i right?
13:18:58markunjhMikeS: because of the new mmu code?
13:19:28n1sTentaclePorn: after inti I think you have to reboot but yeah, after that you should see artists... etc
13:19:51jhMikeSmarkun: I'm betting so
13:20:07jhMikeSthe frambuffer even flashes garbage at startup now
13:20:27 Quit atsea-22 (Read error: 104 (Connection reset by peer))
13:20:28TentaclePornDang, I really needed the database menu too >,<
13:20:47jhMikeSthe calculation of FRAME would be wrong now
13:21:15 Quit mr_pink ("Ex-Chat")
13:22:15*JdGordon hopes whatever jhMikeS fixes will magically fix our mrobe display problem :p
13:23:35n1sTentaclePorn: there are typically two things that break the database build, 1) files that break the metadata scanners for whatever reason (often due to exotic/corrupted/unsupported metadata) or 2) disk corruption so
13:23:53jhMikeScan't say...can't say I care if I break it more either since I don't own one and I need this running now :P
13:24:12n1swhat I would suggest is first a scan of the disk (pretty low chanse it will help but yeah
13:24:23TentaclePornI had a look through and the only files on my iriver are music and the .rockbox folder
13:24:33jhMikeSbut I can say I can't see the point in TTB_BASE being cast to an int * here
13:24:53 Join mr_pink_ [0] (
13:25:00n1sthen trying adding only a part of your music at a time and rebuilding the database to try to narrow down what makes it break
13:25:20TentaclePornyou're a clever man n1s ;]
13:25:42TentaclePornThere is a difference this time around though, it now says "Initialised: Yes"
13:25:44n1sor it you know you have files with odd metadata try without them
13:25:57TentaclePornwill do
13:26:06 Join darksaboteur [0] (i=3aa8784c@gateway/web/cgi-irc/
13:26:17TentaclePornalso, am I supposed to install something new to be able to play doom?
13:26:20jhMikeShmmm...yeah, it might fix it
13:26:38n1sTentaclePorn: yes, see either the manual or PluginDoom wiki page
13:26:43mr_pink_TentaclePorn: I'll second that (about n1s!) - it seems to be building now :) Cheers
13:26:57n1sgreat ;)
13:27:56peturLinusN: did you see my h300 bootloader talk with pondlife this morning (9:30)?
13:28:03jhMikeSback to ok here :)
13:28:08LinusNpetur: nope
13:28:45TentaclePornCheers ;]
13:29:08TentaclePornI just realise, I've been trying to find a place to ask this for a while and this seems to be a perfect place
13:29:26TentaclePornI want to buy a new mp3 player - upgrade from my h10 iriver 20G - any recomendations?
13:29:28markunLinusN: I very much enjoy your blog! (although the swedish is a bit difficult to follow for me)
13:29:28LinusNpetur: ok, so this particular 80gb disk makes rockbox misbehave
13:29:40LinusNmarkun: thx
13:29:55peturonly on my h340 - works ok on the h320 I got recently
13:30:06LinusNpetur: it could be the battery
13:30:16peturit could?
13:30:25LinusNor the pcf50606
13:30:36LinusNit happens when the disk spins up, right?
13:30:41peturmy h340 has a new hi-capacity one
13:31:09LinusNso perhaps it can't deliver the required current to spin up, and the voltage dips
13:31:10 Quit darksaboteur ("CGI:IRC (EOF)")
13:31:28peturI would think that the new 80GB uses less current than the old 40GB
13:31:36peturit certainly makes less noise
13:31:39LinusNyou should try to swap batteries
13:31:43peturboth are dual-platters
13:31:46LinusNas a test
13:32:01peturbah, you know how difficult that is :(
13:32:08LinusNoh yes...
13:32:36TentaclePornNow I have the issue of how to efficiently clean my iriver without having to completely reformat =/
13:32:36peturmaybe I should swap motherboards and we all ignore it ever happened ;)
13:33:07 Join J3TC- [0] (
13:33:10pixelmaLinusN: I second markun about the blog but find it somehow hmm... interesting that it holds much more info than the rest of your haxx page :P
13:33:58LinusNi'm such a lazy guy :-)
13:34:01markunpixelma: but you can read swedish, right?
13:34:34jhMikeSJdGordon: try 'er now
13:34:45 Quit safetydan_ ("Leaving")
13:34:47*petur gets seasick of the blog picture
13:35:12pixelmamarkun: I can understand most of it when reading slowly but still not everything
13:35:37markunpetur: that's what the roads are like over there :)
13:35:49peturand the buildings too?
13:35:58markunand people
13:36:29pixelmadepends on the amount of beer maybe ;)
13:36:30jhMikeSwhich blog pic? wanna see.
13:37:25*JdGordon wonders why the lcd buffer size is 640x480x4
13:37:58 Quit spiorf (Read error: 110 (Connection timed out))
13:38:10 Join XavierGr [0] (
13:38:11jhMikeSJdGordon: that's also goofed? so then it would've been off by 8x its size
13:38:33JdGordonits 1 short / pixel isnt it?
13:39:45*jhMikeS needs a Swedish crash course or to have his gf read it for him :P
13:40:11 Join toolboxx [0] (
13:40:30 Quit toolboxx ("wmIRC(trial)")
13:40:43 Join spiorf [0] (n=spiorf@
13:40:48JdGordonjhMikeS: if all the registers are being cast to a short* wouldnt it make sens the TB_BASE should be cast to a short* also, not a int* ?
13:41:17TentaclePornHey guys, I just put doom on my iriver and I can open the menu and stuff in game but It wont let me select new game or anything o,o?
13:41:52preglowjhMikeS: your gf is a swede? :>
13:42:01jhMikeSJdGordon: it looked like it was cast to (int *) to make the mmy code nicer. in any case the displacement must be watched.
13:42:03TentaclePornnevermind , sorted!
13:42:28 Quit Exiro ()
13:42:40jhMikeSpreglow: no, puerto rican, but she lived in sweden and norway as a student.
13:43:28preglowi'll just not bother with the english with you from now on, then :P
13:44:02jhMikeS:) ... I guess I'll figure something out
13:44:54jhMikeSJdGordon: did that help your FB problem at all?
13:45:13JdGordonnope, no change
13:45:31jhMikeSdid you change it to 640*480*2?
13:45:47JdGordonyeah, just tryed that now, still nothing
13:46:03JdGordonlooks like we are stuck till kkurbjun gets a change to play with it an jtag
13:46:28JdGordoncould it be a stack problem maybe? some lists work and some dont, its really wierd :'(
13:46:59jhMikeSwould they need more stack space on mrobe than usual?
13:47:09JdGordonI wouldnt tihnk so
13:48:02jhMikeSIt sounds like some corruption. maybe make sure all BSS init is correct? I mean _really_ verify it.
13:49:00*JdGordon found FRAME being cast to (int) in the lcd init... that cant be right?
13:49:04 Quit animeloe ("This computer has gone to sleep")
13:49:35jhMikeSwhich file?
13:50:26JdGordonaddr = ((int)FRAME-CONFIG_SDRAM_START) / 32;
13:51:06JdGordonmaybe it is... gcc doesnt like adding a *
13:52:39jhMikeSsubtracting *'s given the number of elements, not bytes
13:53:53 Quit iamben (Read error: 104 (Connection reset by peer))
13:54:23Zagorecho 1 > /sys/bus/usb/drivers/usb-storage/module/parameters/delay_use
13:54:32Zagorrecommended for developers
13:54:34 Join norbusan [0] (
13:54:39 Part norbusan
13:54:54TentaclePornHowdy IRC: Im having issues with getting the database feature to work and Im about to try taking my music off and gradually adding a few songs at a time to see what happens. Should I clean boot, reformat and just reinstall rockbox?
13:54:55Zagorcuts down the delay from connect to device in linux
13:55:30 Join Frazz [0] (
13:55:35Zagoror even use 0
13:57:05JdGordonjhMikeS: that puts addr = 0x1FB300, does that sound ball park ok?
13:57:26 Quit SirFunk (Read error: 110 (Connection timed out))
13:57:56jhMikeSJdGordon: I think so
13:58:28Zagorahh, finally: * read : 170496 bytes in 3001 ms = 56 kB/s
13:58:40Zagorslow, but working
13:58:57Zagorthis is just bulk test code though, not actual ums yet
13:59:01GodEater_TentaclePorn: I suspect using some sort of tagging program to go through and check your music would be quicker
13:59:11GodEater_and just leave the rockbox install alone
14:00:48barrywardellZagor: I read about you're realisations that the USB isn't actually the i.MX31. Now that you mention it, I remember wondering about the number of endpoints before. I found there were three, but just assumed I'd made a mistake somewhere.
14:01:18preglowZagor: woooot!
14:01:28barrywardellthere were other differences too. I think the transceiver type was different than expected
14:01:41preglowZagor: why is that parameter non-zero at all?
14:01:42Zagorbarrywardell: actually I've reversed my claim that it isn't the same as i.mx31. the manual for i.mx31 is very vague about the capabilities, so I can't actually say if it is the same module or not
14:02:08barrywardellso it's a stripped down version?
14:02:13Zagorfor example all the registers say they reset to 000000, which they obviously don't
14:02:43preglowZagor: i don't even have that entry in /sys
14:02:48Zagorno, I rather suspect the manual section is copy/pasted from ARC:s module documentation, which obviously leaves a lot of configuration headroom
14:03:07jhMikeSthis revamped mpegplayer makes some interesting noises anyway
14:03:20Zagorpreglow: it's non-zero because some misbehaving devices don't work otherwise.
14:03:21TentaclePornGodEater: Ill give that a shot, thanks. Any tips as to what sort of stuff im looking for in the tags?
14:03:23barrywardellZagor: lazy of them
14:11:37GodEater_TentaclePorn: I wouldn't bother "looking" for anything. I'd use the tagging program to re-write them all (which should be automatible to some degree)
14:11:46GodEater_that way you know they're all good
14:11:53GodEater_and rockbox shouldn't choke on any
14:13:24amiconnZagor: woohoo, although 56kB/s isn't fast...
14:13:33amiconnIt's in fact slower than the Ondio
14:14:01TentaclePornOhh okay. Which program should I use? Just go through and make sure all their ID3 tags in windows media player are all good?
14:14:07Zagoramiconn: yeah :)
14:14:37GodEater_TentaclePorn: there are a plethora of good tagging programs out there - have a hunt through our forums for recommended ones
14:15:49TentaclePornCheers ;)
14:16:21preglow56kb/s should be enough for everyone
14:17:38TentaclePornGodEater - whereabouts on the forums o,o?
14:17:51LloreanTentaclePorn: Try the "search" feature...
14:17:59*jhMikeS uses mp3tag
14:19:39 Join MethoS- [0] (
14:21:37 Quit MethoS- (Remote closed the connection)
14:22:26 Join Jeton [0] (n=chatzill@
14:24:48 Join agm3nt [0] (
14:28:09*GodEater_ uses easytag
14:28:42 Quit shodanX (Remote closed the connection)
14:28:46 Join shodanX [0] (
14:29:59 Join MethoS- [0] (
14:30:45n1sGodEater_: but this way we'll never find out what it is that kills the database... :(
14:31:55GodEater_n1s: then someone should improve the debugging screen to say which file it got stuck on
14:32:35n1sGodEater_: I thought that was what slasheri's latest commit was supposed to do but it obviously didn't help...
14:33:03GodEater_I don't see anything in the debugging info TentaclePorn posted to indicate which file it was processing
14:33:06GodEater_did you ?
14:35:41 Quit spiorf (Remote closed the connection)
14:39:00GodEater_does hdparm -tT report differences between ipod disk mode and OF connected via usb ?
14:40:27n1sGodEater_: that's what I meant, the currfile thing was added just acouple of days ago to help debugging this stuff but it doesn't seem to work, meh I don't use the thing anyway...
14:41:01GodEater_nor me
14:41:04GodEater_DB is too slow
14:41:35 Join mf0102 [0] (n=michi@
14:43:00LloreanTo me it always seemed like the DB wasn't that useful unless you didn't actually know what music you'd put on your player.
14:43:48n1shaha, I googled for "sansa view dissect" after seeing it on engadget and the first hit was Bagder's blog :-)
14:44:33GodEater_Llorean: that's also a good point
14:46:45Zagorn1s: it's tcc722-based
14:46:52Nico_Pn1s: and it's not a dissect at all
14:47:22n1sZagor: ah interesting, is that similar to the iaudio 7 chip?
14:47:27Zagoruhh, wait I misread that.
14:47:40GodEater_n1s: it will be if Zagor is right :)
14:47:45GodEater_telechips that is
14:51:47 Nick kugel|afk is now known as kugel (i=kugel@unaffiliated/kugel)
14:52:24Zagormeh, can't remember the name of that neat chinese site posting player specs
14:53:35 Quit DataGhost (Nick collision from services.)
14:53:41 Join DataGhost [0] (
14:54:03 Join pixelma_ [0] (i=pixelma@rockbox/staff/pixelma)
14:55:06TentaclePornn1s, do you own a sansa?
15:01:32Zagorah, it is. hard to navigate, but lots of info.
15:06:29 Join midkay_ [0] (
15:06:29n1sTentaclePorn: nope
15:13:43TentaclePornoh okay
15:13:47TentaclePornWhat mp3 player do you use?
15:14:37*JdGordon doesnt think this rtc can count properly!
15:15:22***Saving seen data "./dancer.seen"
15:16:03pondlifeGodEater_: Did your recent wiki update work correctly?
15:16:15GodEater_looked to
15:16:19 Join sup [0] (i=1000@
15:16:27preglowamiconn: cf doesn't have any other tricks for swapping two regs than the good old xor trick, no?
15:16:30GodEater_mainly because I rolled it back again aftewards :)
15:16:40GodEater_when I realised that what I'd added was already there.
15:17:37 Quit newby (Connection timed out)
15:18:40 Quit pixelma (Nick collision from services.)
15:18:40 Nick pixelma_ is now known as pixelma (i=pixelma@rockbox/staff/pixelma)
15:19:28jhMikeSJdGordon: maybe you have to compensate relativic effects?
15:20:24JdGordonno, I dont tihnk so...
15:20:38JdGordonits not following BCD properly, i tinhk
15:21:04jhMikeSpreglow: I saw you q in the logs. Haven't done any more testing on msac. Don't know I said anything to you about that I saw no speed difference? No test_codec runs though.
15:21:55JdGordon22,23,24,28 at the end sorry
15:22:34preglowjhMikeS: don't use it anyway, gas bugs out on msac.w using parallel loads......
15:22:45preglowon that note, are anyone here familiar with the gas source code and could check that out?
15:22:48LloreanjhMikeS: Alsot 20, 21, 22, 20, 21, 22
15:22:50JdGordonthats the order of the first 30s or so got straight from the rtc and put through BCD2DEC()
15:23:03jhMikeSit's i2c based?
15:23:30jhMikeStoo fast a read?
15:24:14JdGordonno, spi is working fine, and this is only being read 1/s
15:24:27 Quit midkay (Read error: 110 (Connection timed out))
15:24:47jhMikeSrunning, the interface too fast wouldn't matter if it's only 1/s
15:24:51*scorche whistles at petur :)
15:25:11JdGordonthe datasheet sucks... it says a heap of bits are set to 0 for writing, but they are obviously not
15:25:13jhMikeSpreglow: msac.w doesn't work with gas?
15:27:22jhMikeShmmm...don't think I ever used it with parallel loads anywhere
15:27:50 Join Darksaboteur [0] (
15:28:05 Join Windows-Server [0] (
15:28:42 Quit Windows-Server (Remote closed the connection)
15:28:45 Quit Darksaboteur (Client Quit)
15:29:01 Join Darksaboteur [0] (
15:29:27preglowjhMikeS: correct, gas messes the encoding up completely, it seems
15:29:47preglowjhMikeS: i went from using msac.w with one constant multiplicand to just neg-ing it and using mac.l, went from locking up to running perfectly
15:29:50jhMikeSyou mean as in invalid instruction?
15:30:11preglownot invalid, it's an msac.w, but the register values are all messed up
15:30:28preglowat least it looked like it when i tried decoding the instruction by hand (head)
15:30:41amiconnpreglow: Swapping how, and which xor trick?
15:30:51preglowamiconn: rX <-> rY
15:31:02preglowamiconn: swapping two registers without using a temp, that xor trick
15:31:29jhMikeSa ^= b, b ^= a, a ^= b?
15:31:37preglowjhMikeS: aye
15:31:57preglowamiconn: i need to do a reverse block copy in qmf_synth() and want to use movem.l
15:31:58amiconnThat works??
15:32:05preglowsure, that's a classic
15:32:07jhMikeSsure does
15:32:40*amiconn didn't know that one
15:32:55preglowit's worth remembering, it's handy here and there :)
15:32:57JdGordonisn using atemp var better? less mem access?
15:33:00amiconnAnyway, you won't save an instruction vs. using a temp reg
15:33:09preglowJdGordon: well, if you have no registers free: no
15:33:15preglowjhMikeS: if you have register free, you save one cycle
15:33:26amiconnpreglow: ?
15:33:29 Join Windows-Server [0] (
15:33:33preglowthat last was to JdGordon ...
15:33:44amiconnUsing temp reg is 3 cycles, and using xor trick is 3 cycles
15:33:59amiconnBut the xor trick works for d regs only
15:34:00preglowamiconn: yes...
15:34:05preglowcompletely true
15:34:33jhMikeSthe xor trick is used in swap32 for ARM
15:34:35*preglow wants something like xchg on x86
15:34:43preglowamiconn: doesn't m68k have xchg?
15:34:53preglowi think perhaps they cut that for coldfire
15:35:21jhMikeSI think they cut a little too much for coldfire
15:36:32preglowyeah, i agree, a wee bit
15:36:37preglowbut it's not too bad
15:36:56preglowi would love the increment/decrement modes fo movem.l to still be there, for instance
15:37:17annulus_jhMikeS, what BCD is that? From is looks like it could be a problem with BCD vs Excess-3
15:37:18preglowthen i wouldn't need any swapping here at all
15:37:41 Quit Sedgewick (Connection timed out)
15:38:17 Nick Windows-Server is now known as Sedgewick (
15:38:39 Nick parafin|away is now known as parafin (
15:38:47jhMikeSannulus_: that's what JdGordon was showing me with what the RTC was producing
15:38:59preglowamiconn: jm confirmed the reason he did the symmetric 16 bit clipping is in case any -32768 values are negated
15:39:13preglowamiconn: so i probably can't use that fast trick everywhere
15:39:20annulus_er right, JdGordon
15:40:41amiconnpreglow: The trick can be adjusted... it'l just a little bit slower in the clipping case, but not in the non-clipping case
15:40:47jhMikeSexcess 3 is still in order, just subtract 3 from it
15:41:03annulus_JdGordon, look at the helpful table for BCD-like encodings
15:41:25preglowamiconn: that's completely ok, non-clipping is the norm
15:41:57amiconnOne extra cycle in the clipping case
15:42:15 Quit Darksaboteur ("ChatZilla [Firefox]")
15:42:57annulus_I was looking at the skip from 4,8 14,18
15:43:08preglowamiconn: how?
15:43:24 Join xazax [0] (
15:44:07JdGordonannulus_: the order its giving me definatly isnt the same as
15:45:44annulus_yeah, it would have had to been the other way around :-)
15:47:32amiconnpreglow: (1) Sub/add 32767 instead of 32768 before/after. (2) Compare with 65534 instead of 65535. (3) After the ext.w, use and.l #65534
15:47:57amiconnWould work best if you could hold both 32767 and 65534 in registers of course
15:48:02 Part xazax ("Leaving")
15:48:59preglowdon't think i have enough regs for the 65534
15:49:15preglowhmm, if i can use an address reg...
15:49:27amiconnNot for the and.l of course
15:49:37amiconnBut for the 32767 you could
15:50:06preglowi already use a data reg for 32767, so i can swap that for an address reg and use the data reg for 65534
15:51:15kkurbjunjhMikeS: did your change fix the problems you were seeing with plugins?
15:51:17preglowat least this'll be easier on arm :P
15:52:43preglowcoldfire allows stuff like move.l (%a0)+, -(%a1), yes?
15:53:05 Join Soap__ [0] (
15:53:07 Quit Jeton ("ChatZilla [Firefox]")
15:53:14preglowwow, and it's fast too...
15:53:56 Quit MethoS- (Read error: 101 (Network is unreachable))
15:56:36Zagorwrite: 1048576 bytes in 3001 ms = 349 kB/s
15:56:38Zagorgetting better
15:56:54LloreanIs there a target speed?
15:57:02preglowi wonder if i should even bother filing a bug report with the gas people
15:57:10JdGordonusb 2.1 speeds :D
15:57:25ZagorLlorean: I'd like to be as fast as OF at least
15:57:25preglowanyone know what retailos transfers at?
15:57:41LloreanZagor: How fast is that, then? :)
15:57:44Zagorbut this is still just test code. speed is not an issue yet.
15:57:53ZagorLlorean: 3-4 MB/s, I think
15:58:08LinusNnot very impressive
15:58:09preglowhope you've got some good hints on what to do, then
15:58:54Zagorgotta go
15:58:56 Quit Zagor ("Client exiting")
15:59:07keanuno! =/
16:00:03 Quit Mouser_X (Read error: 110 (Connection timed out))
16:01:58jhMikeSkkurbjun: all better now
16:02:47kkurbjungreat, thanks for doing that, sorry for the trouble
16:03:50keanuis there a recent patch for USB support?
16:03:53 Join scorche|w [0] (n=8dc5049d@rockbox/administrator/scorche)
16:06:21n1skeanu: <Zagor> but this is still just test code.
16:06:41keanun1s, ok, so no patch yet?
16:07:18n1sjust wait till it's done, 'k?
16:08:08amiconnpreglow: I think I can give you the other d reg back as well (taking an a reg for the 65534 as well)
16:08:40 Quit Soap_ (Read error: 110 (Connection timed out))
16:14:23 Join nicktastic [0] (n=nick@unaffiliated/nicktastic)
16:16:22 Nick midkay_ is now known as midkay (
16:20:38 Join webguest33 [0] (i=39fae588@gateway/web/cgi-irc/
16:20:47 Quit webguest33 (Client Quit)
16:24:20 Quit CaptainSquid ("Miranda IM!")
16:26:15amiconnpreglow: Clipping at the uipper boud would be 1 cycle slower (i.e. no clipping -> 6 cycles, clipping high -> 8 cycles, clipping low -> 7 cycles)
16:26:22preglowhmm, can't the scaled in the indexed addressing mode be negative? :/
16:26:42amiconnNo, only +1, +2, +4
16:27:35preglowwon't that be slower for the no-clipping case?
16:27:50preglowno it won't
16:28:01preglowi'll just go ahead and use that, but no, shop time
16:31:18 Join desowin [0] (
16:37:05 Part LinusN
16:38:14 Join jgarvey [0] (
16:40:37 Quit mr_pink_ (Remote closed the connection)
16:40:45*n1s scratches head
16:44:17n1saaah, stray boost, no wonder it seemed too fast... heh
16:54:33preglowamiconn: the move timings as stated in 5249 manuals are correct as they stand when memory accesses are in iram, yes? even the double transfer?
16:56:29 Join webguest68 [0] (i=d52c79e7@gateway/web/cgi-irc/
16:57:35 Quit webguest68 (Client Quit)
16:58:40preglowargh, is there no way to make proper local labels in asm apart from the numerical ones?
17:01:21 Join Nibbl [0] (n=Nibbler@
17:01:37 Part Nibbl ("Leaving")
17:09:46 Quit z35 (Read error: 110 (Connection timed out))
17:13:30 Join ankka [0] (
17:15:19 Join newbyx86 [0] (
17:15:24***Saving seen data "./dancer.seen"
17:15:35ankkahello... does rockbox nowadays support the 2gen ipod nano or not?
17:16:24LloreanIt says this on the front page of the site...
17:16:30ankkaoh, it said so on the main page after all... the faq isn't very clear on this
17:16:38 Join kkurbjun_ [0] (n=kkurbjun@alamode.Mines.EDU)
17:17:45ankkaIs there going on any development to get it to work on 2nd/3rd gen nanos?
17:18:34ankkaor even the newest ipod classic?
17:18:37markunankka: you can check the forums also for some discussion:
17:18:51markunno even the ipod classic
17:19:20GodEater_which faq isn't clear ?
17:19:34ankkaapple changed the insides of the ipods too much to get it to work anymore?
17:19:46GodEater_they changed it *completely*
17:19:50LloreanI wouldn't call it "changed" so much as "replaced"
17:20:01markunankka: indeed. btw, what did you mean by "it said so on the main page after all"
17:20:28ankkamarkun: I meant that I just missed it on the mainpage
17:20:35JdGordonkkurbjun_: you back?
17:20:40GodEater_ankka - thanks - I'll change that
17:21:02markunankka: it says "(not the Shuffle, 2nd/3rd gen Nano, Classic or Touch)"
17:21:17scorche|wGodEater_: while you are at it, support for 1st and 2nd gen ipods could be added
17:21:23ankka"For iPod: 3G, 4G grayscale, Color/photo, nano and video (5G) and mini 1st/2nd Generation"
17:21:57markunankka: but that's not the main page
17:22:02ankkamarkun: umm yeah I know, I said I missed it and went to read the faq
17:22:12markunah ok
17:22:13ankkaand that the faq was a little misleading
17:22:41scorche|wi wouldnt say it is just doesnt have the NOT clause
17:22:58preglowamiconn: i wonder how a transfer from memory to memory can be just as fast as a transfer from memory to register
17:23:00ankkascorche|w: it does make one think rockbox works on all nanos
17:23:23 Join MethoS- [0] (
17:23:36ankkathanks for your time answering the questions, looks like I'll have to choose between an ipod or rockbox.. :(
17:23:37markunscorche|w: how should you know which ipod nanos and videos are supported or not?
17:23:51GodEater_scorche|w: duh, of course :)
17:23:55scorche|wah...that part, i missed
17:24:03 Part ankka
17:24:05*scorche|w goes off in search of coffee
17:24:22 Quit japc (Connection reset by peer)
17:25:52GodEater_FAQ updated
17:26:07GodEater_I've put in the NOT clause for 2nd/3rd gen Nanos, the ipod classic, and the ipod Touch
17:27:34Nico_PJdGordon: I have a question about the ATA idle callbacks... you remember them?
17:28:02JdGordonyeah, shoot
17:28:31 Quit alienbiker99 (Read error: 104 (Connection reset by peer))
17:28:42 Join alienbiker99 [0] (
17:29:33Nico_PI'm under the impression that they get called very late after the last disk activity, am I right?
17:30:36JdGordonyeah, just as its about to turn off
17:32:22Nico_PI seem to hear the disk turn off then on straight after... isn't that a bit late?
17:32:42Nico_PI was thinking the callbacks would be called as soon as the disk is idle
17:33:01JdGordonthey are called when its ready to shutdown, then turns off straight away
17:34:21JdGordoni cant remember the discussion... but i remember there was one
17:34:24JdGordonyou want it changed?
17:36:28Nico_Pwell I'm relying on this to fill the buffer, and currently the behaviour is a bit weird: buffer the first track, wait, then when the disk is about to turn off, finish buffering
17:36:32markunscorche|w: no problems with the fire I hope?
17:37:04Nico_PJdGordon: ideally as soon as the first track is finished buffering it would finish filling and turn the disk off
17:37:35JdGordonthere isnt really a difference... other than a few secons
17:37:43JdGordonbut maybe the buffering thread should deal with that
17:38:07Nico_Pyeah but the disk is spinning during these seconds... there is energy to be saved there
17:38:24Nico_P(I call ata_sleep() when the buffer is full)
17:38:58Nico_Por maybe I should just manage the initial buffering all by myself
17:39:10JdGordonI tihnk that makes more sense
17:39:54Nico_Pyeah maybe. the callback is to take advantage of a spin caused by the user, is that how I should view it?
17:40:22 Quit petur ("*plop*")
17:40:32Nico_Phmm I haven't phrased that very well...
17:41:02jhMikeSkkurbjun: yes. I became aware of it the minute I tested my revamped mpegplayer. I tested SVN to make sure and many plugins crashed and I suspected that commit.
17:41:06JdGordoni mean, yes thats how you should view the callback
17:41:11Nico_Pok :)
17:42:05 Join webguest63 [0] (i=a98bb464@gateway/web/cgi-irc/
17:42:20webguest63as i understand rockbox uses arm correct?
17:42:27Nico_PJdGordon: is it ok to check whether the disk is active and just start filling if it is?
17:42:37jhMikeSrockbox uses arm?
17:42:42webguest63whats the difference between arm and regular
17:42:48Nico_Pthen maybe I can skip the callback and just do that
17:42:57webguest63i looked on the wiki and didnt see anything about it
17:43:01JdGordonin fact.. you really should check it in the callback if its not a syncronous read
17:43:01jhMikeSregular = coldfire and SH
17:43:13Nico_PJdGordon: which means...?
17:43:15 Quit n1s ()
17:43:19webguest63and arm is?
17:43:21scorche|wmarkun: just hard to would have heard more, but you arent in -community ;)
17:43:32jhMikeSarm = another processor achitechture
17:43:33JdGordonNico_P: like if your callback tells your thread to do something
17:43:59webguest63im using notepad++ what stand alone compiler should i use with it?
17:44:06Nico_PJdGordon: my callback posts a message to my thread's queue... just like in playback.c actually
17:44:19JdGordonyeah, so you have to check ata_is_active
17:44:24Nico_Pthat I do
17:44:28Lloreanwebguest63: You should follow the directions in the wiki to set up the appropriate crosscompiler for your player.
17:44:32jhMikeSwebguest63: there a whole wiki about setting up a build environment.
17:44:47Nico_PJdGordon: but I was thinking of simply checking for that in my main loop and if it's true then buffer
17:44:57JdGordonyou can do that also
17:45:04JdGordonbut it will prolong the disk spinning
17:45:06webguest63ive looked at it but i dont really wanna install another enviroment because my harddrive is getting really low on space
17:45:17JdGordoni.e if your thread catches it before the callbacks are fired
17:45:25JdGordonthen the timeout will restart
17:45:43Nico_PJdGordon: yeah then I don't need the callback anymore... It's for a different use case, isn't it?
17:46:11jhMikeSwebguest63: then it's out of scope on what we usually use. Building is linux based.
17:46:27JdGordonnot really... the buffering should only rebuffer if the buffer is low,it should spin the disk any extra
17:47:09jhMikeSThough VMWare will run on windows and is a 2GB image
17:47:50JdGordonbed time, /me gone
17:47:54 Quit JdGordon ("Konversation terminated!")
17:48:25jhMikeSargh...buffers. /me has to now figure out what's up with the sound stream. the video stream is aok. :\
17:48:27webguest63is there a smaller image that i can use
17:48:49webguest63im only using a 80 gig harddrive and its mostly full
17:49:06Lloreanwebguest63: For bare minimum space use, you'd probably want to use cygwin and determine the bare minimum of packages necessary to install.
17:49:07jhMikeSyou could make one but that's really small. I'd suggest finding something to evict.
17:49:51webguest63why do i get the feeling that its not recomended that i use the bare minimum
17:49:56LloreanjhMikeS: The VMware image is hard to shrink partially because it's reserving that disk space.
17:50:15Lloreanwebguest63: Perhaps because what *is* recommended has already been recommended to you?
17:51:30webguest63my teacher last year had a vitural machine to emulate redhat linux fedora core 2 anyone know of any vms that does that? i remeber it being really small
17:51:59webguest63what if i just used nopix would that work?
17:52:25scorche|wthat would work, but you would have to install the crosscompilers each time
17:53:06 Join A_M [0] (
17:53:14webguest63so basicly i should just grin and bear it and delete some of my stuff to make room for vmware?
17:53:26bluebrotherwebguest63: that would be the best way.
17:53:27scorche|wpretty much
17:53:36webguest63sigh ok thanks guys
17:53:41jhMikeSLlorean: I know, but cywin is too slow for my taste. I want a larger VMWare image since I'm using NTFS anway.
17:53:43bluebrotherand any virtual machine software (like vmware or virtualbox) can run FC2
17:54:15 Quit kkurbjun_ ("leaving")
17:54:22bluebrotheryou can't build Rockbox by just using a compiler. You also need a bunch of other tools −− like perl, bash, etc
17:54:27LloreanjhMikeS: You can resize the drive pretty easily I think.
17:55:10webguest63perl..i thought rockbox was based in c?
17:55:12jhMikeSLlorean: I've no idea how atm.
17:55:40jhMikeSI'm just using a prior player image.
17:56:28GodEater_webguest63: it is - but the build environment uses perl scripts
17:56:40GodEater_amongst other things
17:56:43 Join xazax [0] (
17:56:57LloreanjhMikeS: I may have made a mistake, the solution may just be to add a second disk, I'm not really sure.
17:57:47jhMikeSLlorean: I'll find out when I'm inclined to bother.
17:58:30webguest63wait it uses perl scripts so then just knowing c really wont help with building it will it?
17:59:14 Quit BHSPitMonkey (Connection timed out)
17:59:52 Quit J3TC- (Read error: 110 (Connection timed out))
18:00:27Lloreanwebguest63: Building alone requires knowledge of neither C nor Perl
18:01:14webguest63my appologizes i ment coding it
18:01:27bluebrotherit's the same as with creating an mp3: you don't need to know how mp3 works but you need the tools installed.
18:01:29webguest63u need to have knowledge of both C and Perl
18:01:56bluebrotherno, as long as you don't need to change the perl parts −− which is the build system
18:02:05webguest63i really enjoy rockbox and i am despertly trying to help out the community if only alittle
18:02:05LloreanJust C unless you want to do something advanced or performance critical, then ASM for the appropriate architecture. Perl is only really useful if you want to change how some of the actual compiling goes about its business.
18:02:06bluebrotherthe code itself is C and asm.
18:03:58webguest63asm as in assembly language or automatic storage management?
18:04:38webguest63ok awesome thanks all
18:06:19 Join linuxstb [0] (n=chatzill@
18:09:12 Quit webguest63 ("CGI:IRC (EOF)")
18:20:39 Quit zicho (Remote closed the connection)
18:20:39A_Mhey guys
18:21:02A_Many code gurus around who could perhaps help me out with getting rid of a warning when compiling?
18:21:08A_M"warning: passing arg 1 of 'option_get_valuestring' discards qualifiers from pointer target type"
18:21:43A_Mwhat's it trying to tell me and how do I get around it? (code compiles and seems to be doing what it should btw)
18:22:08 Join J3TC- [0] (
18:24:50A_Moopsie, never mind me, had mistakenly declared the argument I was passing as const *whistles*
18:26:35 Join Bertrix [0] (
18:27:51BertrixI have an iPod nano, I know recording via the line-out isn't going to work(except some stories cranking very high level signals in the port, possibly working due to crosstalk)
18:27:57Bertrixbut will it record via the line-in channel on the dock connector
18:28:06 Join spiorf [0] (n=spiorf@
18:29:12LloreanEvidence suggests it won't.
18:29:34LloreanAt least, I recall people testing the diagnostic mode recording test and getting no results.
18:29:42BertrixIt has been tested?
18:30:21BertrixProbably somewhere on the board will be a trace going to the ADC input, as recording via cross-talk seems possible
18:30:27Bertrixor maybe thats just over the power supply
18:31:24linuxstbI'm pretty sure people have reported line-in recording works on the Nano.
18:32:09BertrixI read that, but I don't know how it can work
18:32:21linuxstbThere is only one "line-in" on any ipod - in the dock connector. Some ipods also have a switchable mono mic input in the right-channel of the headphone socket.
18:33:04BertrixA nano doens't have that right?
18:35:28 Quit Rondom (No route to host)
18:36:57 Quit spiorf (Remote closed the connection)
18:39:01 Join epithetless [0] (i=97c7314f@gateway/web/cgi-irc/
18:41:25linuxstbBertrix: AFAIK, no. At the very least, no official microphones exist for the Nano that attach to the headphone jack (like earlier ipods). Also, there is no code in Rockbox to enable the headphone jack as input on the Nano - it's a GPIO pin that needs to be toggled on the ipods which do support it.
18:41:59Bertrixlinuxstb, ok thanks
18:42:11BertrixIf I have time one day I might try via the dock connector line in pins
18:42:44 Quit qwm (Remote closed the connection)
18:42:55 Join qwm [0] (
18:43:08 Join spiorf [0] (n=spiorf@
18:43:29 Quit epithetless (Client Quit)
18:48:08 Quit Bertrix (Remote closed the connection)
18:51:11 Part linuxstb
18:52:13 Quit shodanX (Read error: 104 (Connection reset by peer))
18:52:21 Join jezzter [0] (i=54d15b74@gateway/web/cgi-irc/
18:53:46 Join japc [0] (n=japc@
18:53:46 Quit jezzter (Client Quit)
18:55:44nomeldoes anyone do firmware development professionally, or are most of you doing this for a hobby/experience points?
18:59:05 Quit J3TC- (Read error: 110 (Connection timed out))
18:59:10 Quit pondlife ("Read error: 110 (Connection slimed out)")
19:00:43nomelwith the usb driver, is it possible to use a device as a soundcard/adc/dac?
19:00:53nomelor, with me c200, as a usb radio?
19:04:07Lloreannomel: Right now it will simply be a UMS device.
19:08:23 Join ilgufo [0] (
19:10:09 Join bertrik [0] (
19:10:25 Quit desowin ("use linux")
19:15:26***Saving seen data "./dancer.seen"
19:15:39 Join Domonoky [0] (n=Domonoky@
19:18:35keanuLlorean, so you're saying it would actually be possible?
19:19:20 Join BigBambi [0] (n=alex@rockbox/staff/BigBambi)
19:19:50 Join Davide-NYC [0] (
19:20:57 Join Rondom [0] (
19:26:25GodEaterkeanu: anything is possible
19:26:38keanuGodEater, nice
19:26:56GodEaterkeanu: however I doubt any core rockbox devs would ever write this as a feature
19:27:02GodEateralthough I've been proved wrong before
19:27:30 Part luckz
19:27:33rasherNot anything.. there are some limitations. For example, it couldn't act as both ums and soundcard iiuc
19:27:58GodEaterI don't see why not
19:28:05 Quit BigBambi (Remote closed the connection)
19:28:09 Quit Davide-NYC ("ChatZilla [Firefox]")
19:28:55rasherGodEater: not at the same time, that is
19:29:02GodEaterI think it could
19:30:03GodEaterI can't think of anything stopping it anyway
19:30:11GodEaterperhaps you explain your reasoning ?
19:30:44 Join J3TC- [0] (
19:31:06rasherWell, I don't quite understand all the details, but I was under the impression that there's a lack of "endpoints", putting a limit to how many things it can do at once..
19:31:38GodEateroh, I wouldn't imagine it would perform well at all :)
19:31:47pixelmaDomonoky: does rbutil pull the newest sansapatcher from the download page or does it use built-in ones, meaning that there could be different versions?
19:31:47 Nick fxb is now known as fxb__ (
19:32:00GodEaterbut if we have proper multi-threading in the core - I think there's nothing to stop it doing both things. It'd just do them badly :)
19:32:09 Join BigBambi [0] (n=Alex@rockbox/staff/BigBambi)
19:32:16GodEaterpixelma: it should use a built in one iiuc
19:32:54preglowamiconn: have you found any timing info on accumulator move instructions?
19:32:57rasherGodEater: I don't think you're right, and that the endpoints thing is a limit that can't be worked around, but I'll leave it to someone with more of a clue to clarify one way or another
19:33:18pixelmaGodEater: hmm, ok. Thank you.
19:33:21GodEaterrasher: fair enough - I'm just going on something Zagor said to me the other day
19:33:51J3TC-rasher: The patch I gave you somehow breaks the compile. I copied straight from the svn diff but something was wrong
19:34:04J3TC-I fixed it tho so here's the working patch for sure...well, tested by me anyway
19:34:38 Quit xazax (Read error: 104 (Connection reset by peer))
19:36:44 Quit A_M ()
19:36:58 Quit miepchen^schlaf (Read error: 110 (Connection timed out))
19:37:13 Join miepchen^schlaf [0] (
19:37:33preglowamiconn: forget that, found it in coldfire2um
19:39:42GodEaterare there any rockboxers in Toronto ?
19:42:19 Join linuxstb [0] (
19:42:59linuxstbGodEater: If only there was some kind of map where Rockboxers could identify themselves...
19:43:37GodEaterlinuxstb: you're not funny
19:43:57GodEateras I just explained in community... I'm really asking "anyone want to go for a beer on Friday, in Toronto"
19:44:02GodEaterwhich the map won't tell me...
19:44:20 Join Davide-NYC [0] (
19:44:28linuxstbApologies for not knowing what was said in -community...
19:44:50GodEaterit should be in your autojoin list ;)
19:45:40Davide-NYCHit and Run idea: test_codec should write the target type and build number to the beginning (or end) of the log file.
19:45:44 Join Buschel [0] (
19:45:59Davide-NYCeasy to do?
19:46:06 Join hcs [0] (n=agashlin@rockbox/contributor/hcs)
19:48:07LloreanDavide-NYC: I can't imagine why it'd be difficult to do.
19:49:13 Quit Davide-NYC (Client Quit)
19:49:26bertrikI noticed that lcd_update for the e200 does a lcd_copy_buffer_rect of the full framebuffer, but there's a TODO that suggests it can be made faster by swapping lcd_driver_framebuffer and lcd_framebuffer
19:49:42bertrikIs this something that I can look into?
19:50:09linuxstbI'm not sure how that could work - they should both be identical after the update.
19:50:36bertrikI'm only summarizing the code comments
19:50:51linuxstbI know - I think I've read that comment.
19:51:09bertrikWho knows about this stuff? Would this make a big speed difference or is there some other bottleneck elsewhere?
19:51:36linuxstb"svn blame" will tell you who wrote the comment.
19:52:31*GodEater guesses dan_a
19:52:56linuxstbAlways someone who isn't around any more... ;)
19:53:14linuxstbThat will be jhMikeS
19:53:17linuxstbWho is around often.
19:53:29GodEaterand whose explanations always makes my head spin
19:54:14rasherWould using floats in a plugin be frowned upon?
19:54:15jhMikeShello :D
19:54:31 Part hcs
19:54:38bertrikhi jhMikeS
19:54:42Lloreanrasher: Isn't the only reason not to use floats performance issues?
19:55:03rasherLlorean: I believe so. In which case it shouldn't be a problem
19:55:05linuxstbrasher: In general, yes. But it depends on the context.
19:55:08jhMikeSbertrik: I think that comment is just old noise. it won't make anything faster.
19:55:29Lloreanrasher: I'd imagine as long as the plugin still functioned "fast enough" on all swcodec targets, you could probably get away with it.
19:56:04jhMikeSGodEater: am I that incomprehensible and/or twisted? :\
19:56:06rasherlinuxstb: a units converter.. a single float*float whenever the user changes the value shouldn't be performance critical, I imagine..
19:56:19amiconnpreglow: Those in COLDFIRE2UM.pdf are for MAC only, not EMAC
19:56:26bertrikjhMikeS: ok, thanks
19:57:07preglowamiconn: yeah, but i couldn't find any other ones, plus, i expect move.l x, %accx and moveclr.l have the same timings
19:57:33amiconnI would expect them to be single cycle on emac
19:57:42*jhMikeS would prefer "twisted" but doesn't get his hopes up :p
19:57:47 Join JETC [0] (
19:58:19preglowamiconn: would sure be a bonus :)
19:59:01 Quit Domonoky (Remote closed the connection)
19:59:24preglowamiconn: i wonder how a transfer from memory to memory can be just as fast as a transfer from memory to register (move.l)
19:59:42amiconnThat I don't wonder about...
20:02:04 Quit linuxstb ("ChatZilla [Firefox]")
20:02:05jhMikeSdual-ported ram? :)
20:02:32preglowi have no idea, it just sounds weird
20:02:40 Quit J3TC- (Read error: 110 (Connection timed out))
20:02:40 Nick JETC is now known as J3TC- (
20:03:37amiconnTo me it doesn't sound weird at all. move.l (mem), reg is 2 cycles, move.l reg, (mem) is single cycle
20:03:53 Join Genre9mp3 [0] (
20:04:37amiconnSo they're 3 cycles in total, but they're 2 instructions. When combining them into one and leaving out the register transfer I would in fact expect to shave off one cycle...
20:05:38 Join ompaul [0] (n=ompaul@gnewsense/friend/ompaul)
20:05:53 Quit amiconn (Nick collision from services.)
20:06:00 Join amiconn [0] (n=jens@rockbox/developer/amiconn)
20:06:35pregloware numerical labels the only real local ones with gas?
20:06:41preglowi thought .labels were local to each major label
20:07:10 Quit japc (Read error: 104 (Connection reset by peer))
20:08:13GodEaterjhMikeS: just much brighter than me I think ;)
20:09:24amiconnrasher: Using floats is frowned upon in general because it's (a) slow and (b) the fp library routines are quite large
20:09:57rasheramiconn: are any of these a real issue in a plugin, provided it's not a performance problem?
20:10:30amiconnOn low-mem targets the fp library routines size may be an issue
20:10:58amiconnThat said, some plugins do use floating point
20:11:04amiconn...e.g. calculator.rock
20:11:33rasherLooks like chopper does as well, which seems unnecessary
20:11:38preglowi'd say it's ok in plugins if avoiding it will be major bother
20:11:46preglowbut if doing without is trivial: do it without
20:11:51amiconnSo chopper should be fixed
20:13:44amiconnHmm, the routines are not as large as I would have thought
20:13:57*amiconn just checked for archos
20:14:38 Quit J3TC- (Read error: 104 (Connection reset by peer))
20:15:03 Join ivan`` [0] (n=ivan`
20:15:14amiconnIn this case, the ones for coldfire are the largest
20:15:24 Join J3TC- [0] (
20:15:30amiconn(quite unusual - usually arm code is larger)
20:15:52preglowperhaps they're hand-coded
20:15:56 Quit ivan` (Nick collision from services.)
20:15:58 Nick ivan`` is now known as ivan` (n=ivan`
20:16:05preglowhand-coded arm asm is usually smaller than compiler generated coldfire code
20:16:17 Join obo [0] (n=obo@rockbox/developer/obo)
20:16:33preglowalso, if rewriting qmf_synth() in asm is not going to give a very decent speedup, i'll eat my shoes
20:16:49*GodEater will pay to witness this
20:16:50preglowhope we'll find out tonight...
20:17:33*amiconn is at 108% realtime for ape -c3000 on coldfire now...
20:17:47GodEaterwell done :)
20:18:25amiconnI hope I can squeeze a bit more
20:19:05 Quit MethoS- (Read error: 104 (Connection reset by peer))
20:20:37 Quit kkurbjun (Read error: 104 (Connection reset by peer))
20:21:39 Join MethoS- [0] (
20:21:40*bertrik is trying to make sense of the LCD graphics API
20:23:06amiconnOne of the few APIs which are actually documented somewhat
20:24:23Buschelamiconn: thank you for moving my patch to .S-file and further sqeezing out of some ms :)
20:24:38amiconnnp :)
20:25:01amiconnI guess you saw that the helper function now allocates the chroma buffer on the stack
20:25:37Buschelyep, as well as using a register for the adress register
20:26:14amiconnThe advantage is that the stack is usually IRAM (for those threads which call it), whilet it doesn't take away precious IRAM statically
20:26:53 Quit MethoS- (Read error: 104 (Connection reset by peer))
20:27:09Buschelnow mpegplayer does nearly reach realtime (no frame drops) for 224x176
20:27:48jhMikeSmpegplayer is a bit fast now without checking actions in the audio loop...which is good I suppose
20:27:58amiconnDid you read about my theory about the finishup() in the log?
20:28:55jhMikeSmy updated version is 4-threaded...still working
20:30:05Buschelamiconn: did you read _my_ tests on LCD-finishup in fs #8014?
20:30:44jhMikeSI think action lists should be sorted at runtime and binary searched. It's quite slow and this registers on a gigabeat.
20:31:02 Join midgey|class [0] (i=8dd53916@gateway/web/cgi-irc/
20:33:42preglowamiconn: move.w never touches the upper part of the reg, righT?
20:33:49amiconnjhMikeS: The action lists have to be checked in their original order to work properly
20:33:58amiconnpreglow: Not entirely
20:34:09keanuIf I do a battery benchmark test for the Sansa, does it matter if the music is being played from a microSD card?
20:34:12amiconnmove.w <ea>, dN doesnt
20:34:24amiconnbut move.w (mem), aN sign-extends
20:34:36preglowok, i'm doing fetches from memory
20:34:41preglowso they sign-extend, ok, that's good to know
20:34:47amiconnInto a-regs?
20:34:49preglowonly address regs
20:35:08*jhMikeS uses that sign-extension in the SPC codec
20:35:12amiconnIt's described in CFPRM.pdf. Refer to movea
20:35:29jhMikeSso we depend on a particular list having a particular order?
20:35:38amiconnBuschel: Hmm, can't find it in the logs by greping??!
20:36:00*jhMikeS thought it looked in the next list if not found in the prior
20:36:06amiconnjhMikeS: Yes. Eralier elements have priority. Otherwise context chaining would never work as intended
20:36:38amiconnYes it does - but the prior list is explicitly allowed to redefine actions for buttons also defined in later lists
20:36:51jhMikeSthat wouldn't change _but_ it could be more intelligent and remember that action _x_ was found in a particular one
20:36:53 Quit miepchen^schlaf (Read error: 110 (Connection timed out))
20:36:59 Join miepchen^schlaf [0] (
20:37:07amiconnDon't overengineer it
20:37:41amiconnThe action list checking is such a small fraction of all code executed...
20:38:21jhMikeSDo I do that?
20:38:56amiconnAnd remembering in which list an action was found would need to be per context
20:40:05jhMikeSI guess just a cache. New context = purge old.
20:40:13amiconnThe whole purpose of this context chaining is to keep the total list size small
20:40:52amiconnIf we would want to waste space, each context would have its own list - but that would a waste of space
20:41:07*amiconn repeats himself in one sentence :/
20:41:35amiconnSuch a waste that would be ;\
20:42:20amiconnIf you get noticeable slowdown from that, you're checking actions too often
20:42:27jhMikeSof course it is. I've never felt any other reason and and not chaning would hardly be better or faster.
20:43:15Buschelamiconn: did not find your theory about finishup in the logs...
20:43:25amiconnNo, me neither :\
20:43:35amiconnI wonder where it went...
20:43:59 Join lazka [0] (
20:44:10jhMikeShaving them in the audio decode loop was definitely checking them too often and now they're only being checked on an actual keypress.
20:44:16Buschelmaybe still between your ears? :P
20:45:42Buschelone of our professors (Prof Dr Sennheiser) always said: "The most of the hearing is done between your ears." Just transferred it...
20:46:18jhMikeSummmm...he's the headphone guy right? :)
20:46:50 Join gryzor [0] (
20:47:00Buschelthe son of the founder
20:47:48 Join shodanX [0] (
20:48:00gryzori have a sandisk sansa e250. I believe it won't be seen as a mass storage device if it runs the original, proprietary firmware. Will rockbox make it easier to access my player from my computer running linux ?
20:48:09amiconnBuschel: starting 02:57
20:49:24amiconnThe 14ms are a strong hint towards that theory. 70fps internal frame frequency...
20:49:32GodEatergryzor: currently the ONLY way to access your sansa over USB is using the original firmware
20:49:43GodEaterand you're wrong - you CAN access it as a mass storage device
20:49:46Buschelamiconn: reading...
20:50:04gryzorGodEater: ty
20:50:22 Quit Rondom (No route to host)
20:50:32amiconnThis needs more experimentation, and probably also disassembling the OF and/or ROM.
20:50:43jhMikeSThe first thing I do when trying a player the first time is go to the options menu. I found the MSC option immediately.
20:54:35 Join tpickers [0] (
20:54:58 Nick tpickers is now known as tpickers66 (
20:55:05 Quit tpickers66 (Client Quit)
20:55:15 Join tpickers66 [0] (
20:56:03 Join Lear [0] (i=chatzill@rockbox/developer/lear)
20:56:12tpickers66can someone upload some new h120/140 perf stats to the wiki, or give me access to do it myself?
20:56:23GodEatertpickers66: what's your wiki name ?
20:57:54 Quit barrywardell (Remote closed the connection)
20:58:42tpickers66GodEater: TimPickersgill
20:59:40GodEateryou now have access
21:00:04 Join xazax [0] (
21:00:56*GodEater goes off in search of food
21:02:21 Quit Genre9mp3 ()
21:06:50 Join freqmod [0] (
21:07:15 Join Domonoky [0] (
21:08:21gryzorGodEater: right now my kernel log does not detect the device as a USB storage. Is there some conf i should set on it that you know of ?
21:09:11*Domonoky made progress with the accessible support in rbutil.. i have now a TabView which is nearly accessible.. but this will take a while to finish.. :-)
21:09:36 Quit Lear ("ChatZilla [Firefox]")
21:10:19gryzorGodEater: found it. Ty for your answers anyway :)
21:13:16 Quit Sedgewick (Connection timed out)
21:15:28***Saving seen data "./dancer.seen"
21:17:11preglowhmm, i forget, do i need to update %sp if i use the stack for local vars and never call anything else? no?
21:20:24 Join Zagor [0] (
21:21:17 Join Honkboy [0] (
21:23:50 Join Nibbl [0] (
21:24:28 Part Nibbl ("Leaving")
21:25:42 Quit freqmod (Read error: 104 (Connection reset by peer))
21:26:40Buschelamiconn: so, you assume something like a framesync. as we always measure about 14ms (independent of the transfer time before −− e.g. 100ms for full screen YVV or 30ms for simple screen update) this would mean we always wait avg 14ms and the LCDs frames all are 28ms? so, the LCD waits for next frame sync to be completed?
21:27:24amiconnSomething like this, yes. As I said - needs more experimentation and research
21:27:46Buschelif so, the theoretical maximum fps is 1000/28ms = 36fps
21:27:53*amiconn thinks the ordinary lcd_update() can be sped up further
21:28:03Nico_PjhMikeS: do you know why messages sent with queue_send aren't received on usb plug?
21:28:04Buschelyes? also tried this but failed...
21:29:36Buschelamiconn: played around with 4byte ldmia and use shifts to set the 2byte registers.
21:29:48amiconnTried stmia for the lcd?
21:29:53Buschelbut i am a beginner to asm :/
21:30:23amiconnThe 0x3000000 / 0x30001000 / 0x30003000 suggests that the lower address bits aren't decoded at all
21:30:38amiconnIn sert "addressing scheme" where you see fit :)
21:32:24 Part Honkboy ("Leaving.")
21:32:41Buschelamiconn: did you try to use stmia?
21:32:57amiconnNo, but I probably will
21:33:10amiconnToo many parallel asm optimisatin projects :/
21:33:15Buschelcan't wait to see your results :)
21:34:38 Join matsl [0] (
21:35:33jhMikeSNico_P: not sure what you're talking about
21:36:55Buschelanother question: could the chroma_buf handling be sped up by using ldr instead of ldrb? so, using 4byte and shifting the needed 8byte out of it?
21:37:01 Quit sup (
21:37:20jhMikeSNico_P: usb_wait_for_disconnect eats all messages until SYS_USB_DISCONNECT is received then it returns
21:37:45Nico_PjhMikeS: ah then my messages are probably sent after the usb connect
21:37:54Nico_Pbecause they all get lost
21:38:25Buschelok, gotta quit. my wife's waiting ;o)
21:38:40 Quit Buschel ()
21:39:42 Join MethoS- [0] (
21:40:01NJoinsup [0] (i=1000@
21:40:24preglowamiconn: do i need to update %sp if i use the stack for local vars and never call anything else? no?
21:40:43amiconnI don't understand what you mean...
21:41:05jhMikeSSpace must be reserved for the locals
21:41:23preglowwhy is that? interrupts? i rememer there was some pitfall here i fell into once...
21:42:08amiconn%sp must be the same before return as it is at the beginning of your function, and you must never write at of above the initial %sp address
21:42:13 Join pixelma_ [0] (i=pixelma@rockbox/staff/pixelma)
21:42:21 Quit pixelma (Nick collision from services.)
21:42:21 Nick pixelma_ is now known as pixelma (i=pixelma@rockbox/staff/pixelma)
21:42:23amiconns/at of/at or/
21:42:39preglowknow that, i'm talking inside functions, if i need to decrement %sp to reflect the actual memory usage
21:42:48amiconnAnd reserving space is necessary not only because of interrupts
21:42:57amiconnSure you have to
21:43:01preglowi never call any other code, so...
21:43:11amiconnOtherwise you would overwrite what your caller has put on the stack
21:43:17amiconn...e.g. the return address
21:43:32preglowamiconn: nono, i am of course allocating properly on the stack, just not updating %sp
21:43:36 Join freqmod [0] (
21:43:36amiconnOr do you mean using negative offsets
21:43:40preglownegative offsets, yes
21:43:48amiconnThat still won't work because of interrupts, then
21:43:58preglowyep, and that was what i was trying to remember
21:44:13preglowi keep forgetting not every arch has separate stacks, like arm :/
21:44:16*jhMikeS hates that lack of an interrupt stack which they fixed of course in a later revision
21:44:22preglowyeah, exactly
21:44:48*amiconn finds those various stacks on arm quite confusing
21:44:55preglowi find them great
21:46:00jhMikeSeach exception mode has it's own banked sp. we just use the interrupt stack for the various aborts too.
21:46:18preglowwe still do? :>
21:46:31preglowi thought i did that as a hack, but i do of course see clearer now :D
21:46:59jhMikeSyeah. It's rare enough why waste the space?
21:48:00jhMikeSshould we set up ARM just like cf? :p
21:48:50Nico_PjhMikeS: what does usb_wait_for_disconnect reply to all the messages it eats?
21:49:15jhMikeSNico_P: the default reply to anything sent is 0
21:50:20jhMikeSthat means if the queue is cleared or a sender is bumped because of overflow or whatever
21:50:25 Quit spiorf (Remote closed the connection)
21:50:50*amiconn *
21:51:10*amiconn *hates* hearing white noise when there should be music :(
21:51:48*jhMikeS hates when he hears white noise when there should be music and find out he's actually listening to a white noise track. :D
21:52:30 Join przemhb [0] (
21:52:44jhMikeSNico_P: perhaps queue_send should reply into a pointer and return true or false? why is 0 a problem?
21:53:15Nico_PjhMikeS: it's not much of a problem, just that I had defined 0 as success so I had to change a few things
21:53:55 Quit nomel ("CGI:IRC (EOF)")
21:56:38 Quit Arathis ("Bye, bye")
21:57:09 Quit xazax ("Leaving")
21:59:32 Join A_M [0] (
22:00:20jhMikeSNico_P: I'd rather modify and improve the interface at this point then fudge around it. It's not used heavily yet.
22:00:54Nico_Pyeah I'm changing the definition of bufclose
22:01:52jhMikeSif it were "bool queue_send(struct event_queue *q, long id, intptr_t data, intptr_t *retval)" then it could be unambiguous.
22:04:50 Quit J3TC- (Read error: 110 (Connection timed out))
22:05:55 Quit mf0102 ("Verlassend")
22:06:27 Quit ilgufo ("So Long, and Thanks For All the Fish -")
22:07:30 Nick idnar_ is now known as idnar (n=mithrand@unaffiliated/idnar)
22:07:56*ender` yawns
22:08:52jhMikeSblah, just found a small bug in queue_delete thanks to looking into it.
22:09:32 Join lee-qid [0] (
22:11:29 Quit homielowe (Remote closed the connection)
22:11:39 Quit przemhb (Read error: 110 (Connection timed out))
22:12:09jhMikeSslimm chance you'll ever see it happen though
22:13:24amiconnAh, no more white noise. :) Silly bug as usual...
22:17:57 Quit amiconn (Nick collision from services.)
22:18:03 Join amiconn [0] (n=jens@rockbox/developer/amiconn)
22:18:18 Join ASCE [0] (
22:18:40ASCEcan rockbox play videos on an ipod video?
22:19:58scorche|wyes, but you would be better off using the OF for that
22:20:45 Part Domonoky
22:20:51ASCEah ic
22:21:03ASCEand what about birightness control on a 5.5g?
22:21:13ASCEcause right now rockbox is killing my ipod :P
22:21:18ASCEipod battery*
22:21:32parafini don't think it's because of brightness
22:21:43parafinand there is a patch for brightness control
22:22:09ASCEim currently looking at that
22:22:14ASCEbut it seems to only be for 5 g
22:22:34ASCEwhat else can drain battery?
22:23:07 Join J3TC- [0] (
22:24:16*preglow expects some white noise about now
22:24:21parafini don't think there is any difference between 5g and 5.5g except hard disk stuff
22:24:44preglownah, just a hang...
22:24:48parafini have 5.5g and patch works very fine
22:24:58ASCEok kool
22:25:11 Join axlgreasetires [0] (i=473ce4bd@gateway/web/cgi-irc/
22:25:16jhMikeSpreglow: why the sudden fit of codec optimizing?
22:25:59ASCEdamn i have to compile
22:26:04ASCEnvm then
22:26:06ASCEthanx for the help
22:26:40 Join gotthardt [0] (
22:26:45ASCEoh one more thing
22:27:01ASCEin the playlist the names come up as ASD or PKE or random stuff like that
22:27:05ASCEwhy is that?
22:27:14Nico_PjhMikeS: amiconn and preglow are in an optimisation competition it seems
22:27:32amiconnAny tester with a gigabeat around?
22:27:39*Nico_P raises hand
22:27:52*amiconn would like to make sure his arm asm opts for ape aren't actually slower on gigabeat
22:28:15amiconn...because gigabeat is arm9, while pp is arm7
22:28:19jhMikeSI guess they all needed it anyway
22:28:31Nico_Pamiconn: where can I find a test track?
22:28:33LloreanASCE: Because apple renamed your files.
22:28:40amiconnInstruction cycle timing is different
22:28:40preglowjhMikeS: because it's fun and i want speex faster :)
22:28:43jhMikeSamiconn: as long as memory stall is avoided, they don't seem too different
22:28:54ASCEso is there any way i can get the right names down in the playlist?
22:29:12amiconnGigabeat is arm920t, right?
22:29:14jhMikeSpreglow: we all want everything faster of course ... keep on keepin' on :)
22:29:24jhMikeSamiconn: yes
22:29:56LloreanASCE: Don't use iTunes to copy the songs.
22:30:12ASCEwould amarock work?
22:30:14 Quit DataGhost ("NTOSKRNL.EXE caused a buffer overflow in System Idle Proce")
22:30:29LloreanASCE: If it's transferring them for use in the original firmware, no.
22:30:40amiconnActually my filter opts aren't worth much on PP *atm*, because even -c1000 is barely realtime and still not usable
22:30:48amiconn-c1000 doesn't use those filters
22:31:12ASCEso any files in the orig. firmware wont have the correct name in playlist?
22:31:21amiconnSomone with a clue needs to optimise the entropy decoder stage :/
22:32:11jhMikeSamiconn: see sec. 12.3. Interlocks. It's the main hoop to jump through.
22:32:16jhMikeSfor 920T
22:32:43amiconnMore than what I thought - on PP5020, -c2000 went from 66% to 73% realtime
22:33:00amiconnThat's just the vector addition and subtraction optimised
22:33:21 Quit axlgreasetires ("CGI:IRC (EOF)")
22:33:23ASCEthanx for the help
22:33:30amiconnI have no idea how to exploit ldmia for the scalar product :(
22:33:35ASCEhaha one more thing
22:33:48ASCEwhen i plug my ipod into the computer, is there any way i can stay on rockbox
22:33:57ASCEcause as soon as i plug in, it restarts
22:33:58 Quit davina (Remote closed the connection)
22:34:25Nico_Pamiconn: are there test tracks available?
22:34:56LloreanASCE: Rockbox doesn't have a disk mode, but holding down Menu prevents it from restarting
22:35:40 Join przemhb [0] (
22:35:42ASCEk perfect
22:35:43ASCEthanx alot
22:36:55 Quit ASCE ("bai!")
22:38:03 Join DataGhost [0] (
22:38:12 Quit bertrik (Read error: 104 (Connection reset by peer))
22:38:28 Nick parafin is now known as parafin|away (
22:38:32 Quit lazka ("I'm off now")
22:40:21przemhbHi all
22:40:23przemhbI've got an issue with incorrectly displayed Rockbox logo and icons. Take a look: It started a few days ago. Now every Rockbox version compiled by me displays logo screen and icons incorrectly. Could someone give me a hint how to solve it?
22:40:51rasherprzemhb: Does the same happen with a build from
22:41:08 Quit midgey|class ("CGI:IRC (EOF)")
22:41:54przemhbrasher: downloaded versions of RB are OK
22:42:20rasherAre you using modified source, or straight svn?
22:42:31 Join petur [0] (
22:42:54przemhbrasher: modified
22:43:12rasherIn which way?
22:43:39przemhbI've added on plugin (my own) and "test_codec" plugin
22:43:51 Join lazka [0] (
22:47:12 Quit nicktastic ("Leaving")
22:48:56 Join japc [0] (
22:49:21 Join bertrik [0] (
22:50:50 Quit matsl ("Leaving")
22:52:16 Quit HellDragon (Success)
22:55:53przemhbrasher: what would you suggest to solve this problem?
22:56:15Lloreanprzemhb: Well, the first thing to do is to make a build with clean, unpatched source and see if it shows the problem
22:58:03amiconnprzemhb: What target?
22:58:09*amiconn guesses H10 big
22:58:23przemhbamicon: yes, you're right
22:58:40amiconn'make clean' and reconfigure. Then rebuild, and it should be fixed
22:59:09 Join webguest35 [0] (i=acce16f1@gateway/web/cgi-irc/
22:59:16amiconnThis is due to the bitmap format change for H10 big (RGB565 -> RGB565SWAPPED)
22:59:31 Join HellDragon [0] (n=Nocebo@unaffiliated/helldragon)
23:02:31 Quit Nico_P (Remote closed the connection)
23:02:40 Join Nico_P [0] (n=nicolas@rockbox/developer/NicoP)
23:03:17 Quit przemhb (Read error: 104 (Connection reset by peer))
23:03:47 Join przemhb [0] (
23:03:49 Quit freqmod (Read error: 104 (Connection reset by peer))
23:05:44 Part agm3nt
23:05:48przemhbamicon: Thank you.
23:06:21 Nick fxb__ is now known as fxb (
23:09:33 Join kkurbjun_ [0] (n=kkurbjun@alamode.Mines.EDU)
23:10:30amiconnHmm, fetching longwords in scalarproduct() might pay off even with the neccessary shifting, at least on arm7
23:12:02 Join linuxstb [0] (
23:14:01 Quit HellDragon (Client Quit)
23:15:29***Saving seen data "./dancer.seen"
23:18:30 Part przemhb
23:21:11 Quit J3TC- (Read error: 104 (Connection reset by peer))
23:21:41 Join J3TC- [0] (
23:22:23preglowhang hang hang
23:23:53linuxstbamiconn: Do you still need a gigabeat tester?
23:24:18amiconnNico_P will test for me, but I have a question regarding libdemac
23:24:42 Quit tpickers66 (Read error: 113 (No route to host))
23:25:37amiconnWhile experimenting with arm optimistations, I found that the first argument of vector_sub() and vector_add() is always 32 bit aligned, but the second argument can be misaligned
23:25:51amiconnFor scalarproduct() it's the other way round
23:25:58 Quit bertrik ("Leaving.")
23:26:07amiconnWould it be "allowed" in your lib to rely on that behaviour?
23:26:18 Quit ompaul (Client Quit)
23:27:03amiconnIt allows for a nice speedup on arm (especially arm7) due to vector_add / vector_sub optimisations
23:27:47 Quit webguest35 ("CGI:IRC")
23:27:50amiconnI also applied that kind of optimisation (just locally so far) to the coldfire scalarproduct(), also giving a little speedup
23:27:56preglowjust look at the code
23:28:04preglowthe first arg is always the start of an array
23:28:09preglowso i'd say yes
23:28:09amiconnI know
23:28:28amiconnI would like to know whether I can rely on this behaviour to stay
23:28:43preglowi'd be surprised if the answer is no
23:28:54preglowbut then i do end up surprised quite a lot...
23:29:10amiconnOn coldfire, breaking that assumption would mean performance drop, on arm it would either crash (data abort) or produce white noise (on targets where aligncheck is disabled)
23:30:48preglowhmm? the fetches fail silently without aligncheck?
23:30:55amiconnSpeedup from the verctor add/sub on arm7: -c2000 66%->73%, -c3000 37%->43%, -c4000 11%->15%
23:31:14amiconnNo, without aligncheck they rotate the word
23:31:30*preglow wonders if we want that behaviour
23:31:49preglowwouldn't it be better to have them fail and consistently find bugs?
23:32:17preglowi guess it can be used for some weird optimizations
23:32:44amiconnddi0100e_arm_arm, page 138
23:32:50amiconn(documentation of ldr
23:32:55 Quit jgarvey ("Leaving")
23:34:06preglowcan coldfire just silently hang if you access illegal memory locations?
23:34:11preglowit surely seems to be what is happening
23:35:14amiconnIf you access empty address areas, you get a bus hang
23:35:22amiconnThe bus cycle simply never completes...
23:35:35*preglow wants an mmu
23:37:08preglowme reset button is probably wearing out as we speak
23:37:24 Quit gotthardt ("ChatZilla [Firefox]")
23:38:57 Quit newbyx86 ()
23:39:41 Quit lee-qid ("aufwiederbyebientotsayonara")
23:39:58linuxstbamiconn: I think that's a fair assumption. But we should probably add some appropriate comments, just in case.
23:41:05amiconnpreglow: btw, ldm behaves different from ldr on unaligned access with aligncheck disabled...
23:41:23amiconnldm just ignores the 2 lsbs
23:42:45linuxstbamiconn: Would it make sense to swap the order of the params to scalarproduct, to keep the alignment requirements consistent?
23:43:30amiconnImho it doesn't matter, but if you want, the swap is trivial, as the scalproduct itself is completely commutatve
23:43:40amiconn(unlike the vector_add and vector_sub)
23:43:56linuxstbI know, but I'm assuming your implementation is assuming the second is aligned?
23:43:58amiconn*scalarproduct even
23:43:58 Quit japc (Read error: 104 (Connection reset by peer))
23:44:09amiconnAtm it does
23:44:21amiconn(for coldfire - arm doesn't care yet)
23:44:46linuxstbThat's all I mean - if you swap it, then all three functions will require the first arg to be aligned.
23:44:53 Join atsea-22 [0] (i=atsea-@gateway/tor/x-acb5da21a247f4bb)
23:47:05 Join donutman25 [0] (n=chatzill@
23:47:15linuxstbDo you have an ARM scalarproduct yet?
23:47:35amiconnFor arm9 it probably doesn't pay off to load 32 bit packed
23:48:03amiconn..because arm9 ldrh is single cycle, unless the result is needed in the next instruction
23:48:13amiconnBut on arm7 ldrh is 3 cycles...
23:48:31 Quit petur ("here today, gone tomorrow")
23:48:41linuxstbWhat is a normal ldr?
23:48:45linuxstb(on arm7)
23:49:25amiconnAlso 3 cycles
23:49:50amiconnBut (1) this means loading 2 values at once and (2) loading 32 bit words allows using ldmia
23:49:52 Join kx [0] (n=kx@
23:50:01amiconnAnd ldmia is (n+2) cycles
23:50:33amiconnSo loading 4 32 bit words with ldmia needs 6 cycles, while loading 8 halfwords with ldrh needs 24 cycles...
23:51:33amiconnLooking at that, it even pays off to do some bit shuffling afterards
23:52:17 Quit Frazz (Read error: 110 (Connection timed out))
23:52:51 Join animeloe [0] (n=animeloe@unaffiliated/animeloe)
23:53:00 Quit obo ("bye")
23:56:04 Part kx ("Leaving")
23:56:19 Join stripwax [0] (
23:57:43 Quit stripwax (Client Quit)
23:58:02 Join stripwax [0] (
23:58:13 Join japc [0] (

Previous day | Next day