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

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

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

#rockbox log for 2005-03-02

00:00:56 Join markun [0] (
00:01:29LinusNpreglow: system_init() in system.c would be a good place
00:01:48[IDC]DragonLinusN, amiconn: any opinion about Jerry's new V1 charging?
00:02:13LinusNpreglow: i did some thinking, and i think the current burst mode of the SDRAM might be the optimal methid after all, instead of page mode
00:02:16amiconnI did not yet check it out.
00:02:24LinusN[IDC]Dragon: see my response in the mailing list
00:02:26preglowLinusN: why?
00:02:39[IDC]Dragonoops, rtfml
00:03:06LinusNbecause the code will be in internal ram, so there will always be interleaved sram/dram fetches
00:03:20amiconnLinusN: Speaking about the ml - I still get the monthly reminder twice
00:03:23[IDC]DragonRead The F***ing Mailing List
00:03:27LinusNthus the sdram page will be closed on each sram fetch
00:03:47LinusNi get it twice too, it's an old mailman bug, still unfixed
00:03:55amiconnLinusN: Nothing similar to SH1's RAS down mode?
00:04:06preglowLinusN: what would be the disadvantages of turning it on?
00:04:12LinusNamiconn: it's almost exactly the same thing
00:04:27[IDC]DragonLinusN: combine trickle/deep into one option?
00:04:42amiconnIf there is such a mode, enabling page mode should boost performance...
00:04:44LinusNa page mode "burst" is about 5-4-4-4...
00:04:54LinusNa real burst is 5-1-1-1
00:04:58preglowthen i agree
00:05:04 Join webguest25 [0] (
00:05:36LinusNamiconn: the coldfire can't do internal accesses simultaneously as external ones
00:05:42 Quit webguest25 (Client Quit)
00:05:44 Join midk [0] (
00:05:52LinusNno delayed writes either
00:06:14LinusNextremely basic bus controller imho
00:06:43amiconnHowever, RAS down mode has nothing to do with warp mode (simultaneous internal & external accesses)
00:06:57preglowand while we've got a 68k person here, can you comment on this, linus? :
00:06:58preglow20:34 < preglow> gcc doesn't seem to think the coldfire can compare with a constant, or is there some other reason for it always loading the constant to another register for comparing?
00:07:01LinusNso the only way to take advantage of burst mode for data access is using movem a lot
00:07:31amiconnRAS down mode merely tries to hold the page open as long as possible, i.e. also when the SH accesses external areas outside the RAM (I/O space, ROM etc)
00:07:42LinusNpreglow: what was the type of the variable that it compared the constant to?
00:07:56preglowLinusN: int
00:08:21preglowLinusN: it did fit in a movq in that particular case, but i've seen it do it for bigger numbers as well, and then with an ordinary move
00:08:31LinusNamiconn: yes, but the coldfire bus controller doesn't keep it open that long
00:08:50LinusNpreglow: weird indeed
00:09:19LinusNi think the coldfire gcc implementation can get better
00:09:33preglowLinusN: i'm trying to familiarise myself with writing whole functions in asm by reoptimizing the flac thing, and i bumped into it while comparing an iterator
00:09:43preglowyes, so doi
00:09:49preglowit does some strange things from time to time
00:10:01LinusNpreglow: was the iterator in a data or address register?
00:10:10preglowLinusN: data
00:10:19preglow 31c: 7008 moveq #8,%d0
00:10:19preglow 31e: b085 cmpl %d5,%d0
00:10:45 Join amiconn_ [0] (
00:10:45LinusNthat's faster that a long compare
00:10:57preglowlong compares are single cycle
00:11:03preglowand that is a long compare
00:11:09preglowthe coldfire only has long compares
00:11:11LinusNyes, but the opcode is 3 words
00:11:25stripwaxIf you're doing DMA from sdram only, will turning off the 'dma from sram enabled' bit speedup DMA (since it can bypass the sram hit logic)? just an idea
00:11:26 Quit amiconn (Nick collision from services.)
00:11:27 Nick amiconn_ is now known as amiconn (
00:11:44preglowLinusN: yes, but i'd say it's better to optimize for speed than space in this particular case
00:11:46LinusNstripwax: no
00:11:59LinusNpreglow: the fetches are not for free
00:12:01preglowLinusN: that code will be two cycles, as compared to a cmp.l with constant, which should be one cycle
00:12:30preglowLinusN: fetch? the const is encoded in the instruction
00:12:31markundamn, ogg's decode at 54% real-time here!
00:12:45preglowLinusN: or do you mean code fetch from ram here?
00:12:48stripwaxmarkun - neat, that's quicker than mp3s
00:12:50HClpreglow: it needs to do 3 fetches for a 3 word instruction...
00:12:51LinusNcmp.l with constant is 3 words, which means 3 memory accesses to fetch the instruction
00:13:00markunI disabled the printing of the used space in malloc..
00:13:18preglowtouch luck anyway, i haven't got any registers to spare :PP
00:13:33*HCl knows that feeling
00:13:38LinusNbut moveq/cmpl is two words to fetch
00:13:49HCland to execute?
00:13:55amiconnmarkun: I also thought about this. It isn't necessary to disable this, it could go into the "update at most once per second" section instead
00:13:57LinusNthe external ram is 16 bits btw
00:13:59HClhow many cycles for that..
00:14:00preglowLinusN: i will use it if i've got registers to spare
00:14:20preglowLinusN: and btw, a nop is three cycles, not one, like you comment in lcd.S
00:14:38markunLinusN: I also wanted to do this, but I first tested without printing to see how much faster it is.
00:14:42preglowLinusN: you've got the math right, seems, just not the comment
00:14:44LinusNSH1 thinking...
00:15:08markunAnother this I changed is a bigger buffer and I completely fill the buffer before I send it to disk.
00:15:37markunNormally it only decodes 4096 bytes (1 frame)
00:15:42HClwow, a nop is 3 cycles?
00:15:48HClthats rather expensive for a nop...
00:15:49preglowHCl: yes, it syncs the pipeline
00:16:03preglowHCl: there is a single cycle one as well, just don't remember what it's called
00:16:10HClit makes sense then.
00:16:24HClhas anyone tried rockboy with the speed increase?
00:16:31preglowHCl: nope
00:16:33HClotherwise i will after my shower.
00:16:51HCldon't need to update the bootloader for the speed increase thing, right?
00:16:52DMJCso what's the latest news?
00:16:54markunHm, another ogg only decodes at 34% :( The first one I tried was mono.
00:16:56preglowHCl: nope
00:17:05preglowmarkun: at 96mhz?
00:17:06HClDMJC: linus comitted his speed increase stuff..
00:17:11HClok, bbl, showerrr
00:17:12markunpreglow: yes
00:17:14DMJCwhat clock speed?
00:17:16preglowmarkun: that's not bad at all
00:17:34DMJCor less?
00:17:47preglowyou can still get 11 as well
00:17:59DMJCI'll build/test it now
00:18:09 Quit Dicko (Remote closed the connection)
00:18:23DMJCthere's no way to actually get sound output is there?
00:18:24LinusNpreglow: have you tried to put some code and data in iram?
00:18:48preglowLinusN: nope, spent a good three hours walking around in the woods and have been doing flac asm since that
00:18:56HClwoods are great.
00:18:59*HCl nods.
00:19:01LinusNpreglow: oki
00:19:03*HCl bbl now
00:19:08preglowHCl: indeed, and the weather was fantastic here today
00:19:12preglownice and cold
00:19:14HCleh heh.
00:19:18HClit was icey here today
00:19:22DMJCLinus, did you try 140 mhz, and if so what happenned?
00:19:24HCli completely slipped with my car into a parked car
00:19:26HCl :X
00:19:28preglowtons of snow
00:19:32HClbut thats what insurance is for.
00:19:33preglowi don't have a car, so i love snow
00:19:52LinusNDMJC: it crashes after a while if the hard drive spins too much
00:20:04LinusNmy guess is that it gets too hot
00:20:30LinusNperhaps not a real world case, but i don't want to push our luck
00:20:39preglowLinusN: i haven't looked much at it, but all you do in mpa2wav is a single memcpy to move stuff to iram?
00:21:12preglowLinusN: so i just need to mark more stuff as belonging in iram and it'll work?
00:21:29preglowvariables will attend to themselves i hope?
00:21:32preglowafter i mark them
00:21:40LinusNunfortunately, you have to change lots of the tables to non-const
00:21:58LinusNvariables will "just work"
00:22:14LinusNbecause the .idata section is non-const
00:22:26preglowahh, like that
00:22:41preglowi thought const was a pure language level thing
00:22:53preglowbut nevermind
00:24:31preglowwould anyone happen to know how i make objdump eat a plain binary file?
00:25:15 Quit methangas (" HydraIRC -> <- 100,000+ downloads can't be wrong")
00:25:36LinusNpreglow: −−target=binary
00:25:56preglowahh, i tried wrong parameter, thanks
00:26:13LinusNbut you might need to specify −−architecture as well
00:26:13 Quit lolo-laptop ("Client exiting")
00:27:21preglowyes, so it seems
00:28:02preglowdoesn't work
00:28:12LinusNwhat are you trying to do?
00:28:27preglowdump a plugin to see if i've gotten it to link properly
00:28:30preglowi use two .o files
00:28:40preglowto make one plugin, that is
00:29:02amiconn[IDC]Dragon: Re your commit message... the icon was created by my sister.
00:30:05LinusNpreglow: m68k-elf-objdump −−target=binary −−architecture=5200 -DS myplugin.rock
00:30:38preglowok, i thought m68k would do for an arch
00:30:51preglowworks excellent, thanks
00:31:26amiconnLinusN: How could I check for 2 conditions to be true (AND) in a Makefile?
00:31:28[IDC]Dragonamiconn, then credits for you showing me ;-)
00:32:00amiconnI mean, doing one thing when both conditions are true, else another thing
00:32:45LinusNnot sure
00:32:58 Part [IDC]Dragon
00:34:08preglowit actually looks like it isn't linked in...
00:34:16preglowhow the hell is that possible without it complaining
00:36:48HCldo i need a debug version for the cpu scaling?
00:37:11HCljust checking since it was under debug options, right..?
00:38:18DMJCwhat is .rockbox/viewers/vorbis2wav.rock
00:38:34 Join AC [0] (
00:40:35ACif i run make zip, the zip files only contains rockbox.iriver.. no .rockbox folder
00:40:49preglowDMJC: it's the vorbis 2 wav converter?
00:41:05preglowAC: you sure you haven't made just the bootlaoder?
00:41:29ACi am sure
00:41:38ACi have run ../tools/configure
00:41:40LinusNAC: how do you see that the .rockbox folder isn't there?
00:41:46DMJCwhat I meant is, does rockbox have to convert the audio to wav to play it back?
00:42:01LinusNDMJC: it doesn't play it at all
00:42:16ACLinusN: I have opend the zip file :)
00:42:30LinusNAC: which operating system?
00:43:02ACoh found my mistake.. kde hides all folders beginning with a dot
00:43:04DMJCwhere do I go to change cpu clock?
00:43:12LinusNAC: suspected that
00:43:23LinusNDMJC: debug->view i/o ports
00:43:42LinusNthe UP for 96MHz, DOWN for 48MHz and SELECT for 11MHz
00:44:27DMJChow will I know if it's working?
00:45:10preglowDMJC: try decoding something before and after
00:45:24LinusNit says so, plus you will notice a speed increase in all operations
00:46:17DMJCso snake will be much faster/
00:47:41preglowbut argh
00:48:55DMJClol at chicago 12
00:49:10preglowbut ok, i've stuffed clearing of accumulators in system_init, i'll assume they're always cleared from now on
00:49:15ACwavpack decodes at about 20 % :(
00:49:16DMJCscrolling is much niceer
00:49:21ACat 96 Mhz
00:49:35preglowAC: then time to optimize!
00:49:46 Quit stripwax (Read error: 110 (Connection timed out))
00:50:02DMJChow the heck do you optimize this stuff?
00:50:09preglowDMJC: yes
00:50:14ACpreglow: yep.. i will profile the lib and look what functions are call often..
00:50:22preglowAC: good
00:51:15preglowread up on coldfire asm + emac instruction, and get ready for fun
00:51:28amiconnAC: I tried wv2wav in the simulator with your test file. It says "crc errors", and then the .wav is not playable... (of course, since the correct wav header doesn't get written)
00:52:02ACamiconn: tryed id some seconds ago.. works here fine
00:53:36amiconnIn the simulator?
00:53:44ACno.. on the target
00:54:28 Quit Sucka ("a bird in the bush is worth two in your house")
00:57:48ACmaybe Crossware Coldfire Development Suite is worth for trying?
00:58:03preglowAC: for what?
00:58:34ACoptimizing the codecs, because it has an simulator in it
00:58:55preglowthought you meant using the compiler
00:59:06ACno :)
00:59:14preglowthat would be fine
00:59:30ACyep.. i will try the trial
01:00:55 Join Camilo [0] (
01:02:51amiconnHCl: r u around?
01:02:52***Saving seen data "./dancer.seen"
01:02:53preglowhaving a simulator would be really Grand
01:03:15DMJCholy crap
01:03:26DMJCanyone see the THG review of the 6800 go ultra?
01:04:54preglowLinusN: no damage in me using the frame pointer for other stuff?
01:07:57 Quit Stryke` ("Friends don't let friends listen to Anti-Flag")
01:10:33ACgood night all
01:11:12 Quit AC ("CGI:IRC (EOF)")
01:11:37HClamiconn: yea
01:12:15preglowLinusN: do you have time to take a quick peek at a small asm function i'm writing to see if i'm doing something fundamentally wrong?
01:12:16amiconnI just checked the adapted sokoban on archos. The walls are all black now :(
01:15:21LinusNpreglow: why would you?
01:15:57HClyes, they are o.o
01:16:02preglowLinusN: i've never written a 68k asm function before, and it's the first time i use jump tables in asm :)
01:16:16HClits a) more efficient lcd wise, and it was also the only easy way to make them, i made any outside white
01:16:33amiconnHCl: The black walls make it harder to see your player "block"
01:16:40HCl :X
01:16:50HClokay, well, i tried to make a quick algorhythm that was size independant
01:16:54HClbut i didn't really succeed in taht.
01:16:59HClso i just made them black
01:17:12HClyou're free to improve that, but i still suggest keeping the outside white cause it saves battery
01:17:19HClat least, i think it does.
01:17:33HClblack = on = needs lcd to be black = needs more power?
01:17:38LinusNpreglow: the easiest way is to write a stub in C and see how it looks
01:17:41HClor am i wrong in thinking that...
01:17:53amiconnThe display contents of an lcd doesn't influence battery consumption at all
01:17:55preglowLinusN: yeah, done that
01:18:02HClhm, sure?
01:18:06preglowamiconn: it does, but not much
01:18:14*HCl nods.
01:18:19LinusNpreglow: i have to go to sleep, but i can look at it in the morning if you like
01:18:24 Join lostlogic [0] (
01:18:27preglowLinusN: no hurry
01:18:35amiconnpreglow: If you consider a few ľA significant, then yes
01:18:36LinusNif you can't get it to work, just send it to me
01:18:56preglowLinusN: again, no worries, i'll probably figure it out
01:19:03LinusNwell, nite all
01:19:05preglowgood night
01:19:09 Part LinusN
01:19:11preglowamiconn: no, it is insignificant
01:19:27preglowlcd power usage is one of the things you can safely ignore
01:19:31pregloweven if it's colour
01:20:09amiconnAn lcd cell is an isolator by definition
01:20:14HCli didn't know that :)
01:20:33amiconnpreglow: But with a colour lcd, there's the backlight problem
01:21:07preglowamiconn: yup, and that is significant
01:21:14preglowvery much so
01:21:25amiconnHCl: The white outside area actually looks better than before, imho. Only the black walls are irritating
01:21:48amiconn*are actually looking
01:21:51HClfeel free to fix it :)
01:22:34 Join chuck [0] (
01:22:46amiconnA 50% dither shouldn't be hard to get size indendent. It will look a bit odd if the block size is not a multiple of 2 though
01:23:39amiconnAh, no, it's even possible to make it look good in that case
01:24:03amiconn<HCl> why..? <== because I'm stupid (sometimes)
01:24:08*HCl grins
01:24:10HClno worries
01:24:14HCli'm stupid too :p
01:25:49preglowcount me in!
01:28:51mstAnyone would have any documentation on Sigmatel STMP34xx chips?
01:32:43 Quit Camilo (Read error: 110 (Connection timed out))
01:44:35amiconnHCl: My new Makefile stuff is working.
01:45:03amiconnI also killed those 4 nasty warnings. rockboy got 2 kbytes smaller...
01:45:32amiconnThis is finally an example why having code in a .h file is a bad thing
01:45:32preglowamiconn: when will we see this submitted?
01:48:56 Quit markun ()
02:03:40HClnice :)
02:03:47HClamiconn: yup
02:04:03HClamiconn: did you add dynarec? with it turned off by default?
02:04:15amiconnNo, didn't add it yet
02:04:45HClcould you?... i really prefer the dynarec version to be in cvs cause it has some changes and the general framework...
02:05:02HCljust cause i don't want two branches of the code.. mostly
02:05:10HClone in rockbox cvs and one in my local cvs
02:06:08amiconnI'm not going to commit this just now, perhaps tomorrow. Some little things are still left to be done, and I need to sleep....
02:06:18HClno need to hurry.
02:06:35HCli'm gonna sleep too.
02:06:45HCli promised my gf i'd go to bed early, and its already 2 am :(
02:06:47HClso, night.
02:21:15 Part amiconn
02:31:22 Join geoff_o [0] (
02:33:28geoff_oDoes stripwax hang on on irc?
02:41:20preglowyes, but not right now
02:45:43 Join DjMnG [0] (
02:47:25 Join condor9 [0] (
02:51:24condor9I'm trying to flash 2.4 rombox.ucl (or rockbox.ucl) but when I play it, nothing happens.
02:52:20condor9I'm able to flash the initial firmware ... but it doesn't seem to recognize the ucl files.
02:52:24condor9Can anyone help?
02:55:41 Quit condor9 ("Leaving")
02:57:42 Quit cYmen ("leaving")
03:02:55***Saving seen data "./dancer.seen"
03:18:22 Quit preglow ("leaving")
03:31:45 Quit Aison ("( :: NoNameScript 3.72 :: )")
03:37:20 Quit mecraw ("Trillian (")
04:05:35 Join QT_ [0] (as@area51.users.madwifi)
04:09:21 Quit chuck ("thanks, jerk.")
04:14:07 Quit QT (Read error: 60 (Operation timed out))
04:39:42 Join condor9 [0] (
04:39:59 Quit condor9 (Client Quit)
04:41:58 Part geoff_o ("Kopete 0.9.1 :")
04:44:27 Quit ze (Read error: 104 (Connection reset by peer))
04:48:34 Join ze [0] (
05:02:57***Saving seen data "./dancer.seen"
05:10:29 Quit uncledrax ()
05:27:32 Join webguest20 [0] (
05:27:58 Quit webguest20 (Client Quit)
05:28:01 Join webguest20 [0] (
05:32:16 Quit webguest20 (Client Quit)
05:46:22 Quit Ka_ (Read error: 110 (Connection timed out))
05:47:20 Join ze__ [0] (
06:03:16 Quit ze (Read error: 110 (Connection timed out))
06:03:17 Nick ze__ is now known as ze (
06:03:23 Quit nozomiyume (Ping timeout: 14400 seconds)
06:33:02 Join ashridah [0] (
06:55:55 Quit midk ("Leaving")
07:03:00***Saving seen data "./dancer.seen"
07:20:21 Join StrathAFK [0] (
07:38:55 Quit Strath (Read error: 110 (Connection timed out))
08:07:44 Nick StrathAFK is now known as Strath (
08:17:05pabsError: unrecognized architecture specification `5249'
08:17:36pabsi've installed the latest binutils from cvs (both for the m68k-elf, and native), and the latest gcc tarball
08:17:39pabsi guess i'll try gcc cvs
08:20:55pabspabs@halcyon:~/cvs/rockbox> which nm && nm −−version | grep nm && which gcc && gcc −−version | grep gcc
08:21:00pabsGNU nm 2.15.95 20050302
08:21:05pabsgcc (GCC) 3.4.3
08:21:07pabspabs@halcyon:~/cvs/rockbox> gmake
08:21:07pabsgmake[1]: Entering directory `/home/pabs/cvs/rockbox/firmware'
08:21:07pabsCC backlight.c
08:21:07DBUGEnqueued KICK pabs
08:21:07pabsAssembler messages:
08:21:10pabsError: unrecognized architecture specification `5249'
08:21:12pabsgmake[1]: *** [/home/pabs/cvs/rockbox/backlight.o] Error 1
08:22:05pabspabs@halcyon:~/cvs/rockbox> m68k-elf-gcc −−version
08:22:05pabsm68k-elf-gcc (GCC) 3.4.3
08:22:24pabswhat am i missing here?
08:24:24 Join crwl [0] (
08:25:46 Quit shx ("CGI:IRC (EOF)")
08:26:04ashridahare you using cvs binutils?
08:26:18ashridahnevermind, didn't look up
08:26:33ashridahbeats me. i think i've seen it before tho mentioned by someone else. search the irc logs?
08:26:47pabsi did already, that's how i ended up here
08:27:22pabsi just sat here recompiling binutils and gcc several times to get teh combination right
08:27:52pabswait a second here
08:28:38pabsteh instructions say i only need cvs gcc-3.4 for calmrisc, is it possible i need it for coldfire as well?
08:30:33Strathcalmrisc16 isn't supported in the official gcc release, i don't know about "coldfire"
08:31:16pabsi'm going by these instructions"
08:32:22Strathwe had to write our own gcc backend (with some help from samsung:
08:35:49pabsah well, ij ust want to build the simulator anyway
08:35:53pabslooks like that's compiling
08:35:55Strathfrom a google search, it looks like the coldfire backend has already been submitted and added to gnu gcc
08:38:18Strathand judging by that page in the rockbox wiki, it also looks that way ;)
08:40:37Strathheh... m68k-elf, easy enough :D
08:43:35ashridahpabs: i only used cvs binutils, and gcc 3.4.2 's -core release
08:52:06CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
08:52:06*pabs wonders if audio works from the simulator
08:53:20pabsi'll take that as a no
08:53:36 Join LinusN [0] (
08:54:23LinusNbrrrr, it's -20 degrees celsius today
08:54:52ashridahhaha. and i was bitching about the cold change that made it dip to 19 degrees here :)
08:57:18dwihnoSomeone must've installed a gigantic peltier element
08:57:51ashridahi'd hate to be on the other side of that peltier
08:58:23dwihnoI just got a phonecall. Special reindeer delivery ;)
08:59:00LinusNpabs: do "m68k-elf-as −−version"
08:59:52LinusNseems you have used binutils-2.15
09:00:00LinusNand not cvs
09:03:01***Saving seen data "./dancer.seen"
09:03:05pabsLinusN: grarr, how the hell did that happen?
09:03:13*pabs starts over again
09:03:28LinusNyou must have had binutils-2.15 in the path when you built gcc
09:03:31pabseverything else is cvs
09:04:09pabsyou know what really sucks?
09:04:21pabsi built the native ar correctly
09:04:24pabscheck this out:
09:04:32pabspabs@halcyon:~/src/m68k-toolchain> ar −−version
09:04:33pabsGNU ar 2.15.95 20050302
09:04:46pabsas opposed to:
09:04:47pabspabs@halcyon:~/src/m68k-toolchain> /usr/bin/ar −−version
09:04:47pabsGNU ar 2.15
09:05:07pabsah well, at least i know what's messed up now
09:05:10pabsLinusN: thanks
09:05:39pabsLinusN: now where's my h120 audio?
09:05:55pabs(as in, you're supposed to be working on that)
09:06:05LinusNoh, am i? :-)
09:06:59pabspabs@halcyon:~/src/m68k-toolchain/build> m68k-elf-ar −−version
09:06:59pabsGNU ar 2.15.95 20050302
09:07:00pabsthere we go
09:07:44pabsit's from cvs
09:08:08pabsam i using the wrong repo?
09:08:19LinusNmine says: GNU ar 050217 20050217
09:08:41pabsmine's still givingm e the date
09:08:53LinusNgrab the tar snapshots from their ftp server
09:09:13pabswell lemme see if this works for me
09:09:16pabsif it doesn't then i'll dot aht
09:09:49pabsyup this works now
09:10:32pabsLinusN: this is compiling fine now
09:15:31LinusNi have updated the crosscompiler page about the cvs snapshots
09:17:27pabsLinusN: k
09:17:35pabsLinusN: you might put something up there about checking hte versions
09:17:41pabsLinusN: and the error message i got
09:18:02pabsLinusN: that's showed up in irc at least 4 different times (i know you helped someone with it once)
09:20:56 Join jyp [0] (
09:27:56 Nick kergoth is now known as kergoth`zzz (
09:39:46jypLinusN: gratz for the changing frequency :)
09:40:24DMJChehehe "boost"
09:40:34DMJCwhy do I get visions of irivers on nitro
09:40:42LinusNjyp: you should look at the boost API and do that for the gmini too
09:41:01jypA quick question cause I'm lazy... What's the define to check for cygwin (unrelated to rockbox)
09:41:22LinusNDMJC: we was about to call it "turbo", as a tribute to the "turbo" button on old 386 PC's :-)
09:41:44jypLinusN: sure; I'm eager to look into it
09:41:48 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
09:42:10 Join einhirn [0] (
09:42:57ashridahLinusN: so we're sticking with 96MHz as the max frequency?
09:43:55LinusNjyp: i believe it's __CYGWIN__
09:44:04LinusNashridah: for now, yes
09:44:29jypLinusN: thanks many
09:44:33LinusNwe might raise it later on, when we use the boost API for real
09:44:42 Quit ashridah ("phone")
09:45:16LinusNi have to make some more measurements to make sure we don't fry any cpu's
09:50:31jypAlright... I'll be careful too then ;)
09:51:11jypThe gmini SP doesn't seem to heat _at all_ when running the Archos firmware
09:57:01 Nick Lynx_awy is now known as Lynx_ (HydraIRC@
10:13:32 Join webguest45 [0] (
10:13:51 Join ze__ [0] (
10:22:34 Quit Nibbler (Read error: 54 (Connection reset by peer))
10:24:03 Join ashridah [0] (
10:27:56 Join preglow [0] (
10:28:14 Join Nibbler [0] (
10:28:14 Join NibbIer [0] (
10:28:32 Quit NibbIer (Read error: 104 (Connection reset by peer))
10:30:12 Quit ze (Read error: 110 (Connection timed out))
10:30:18preglowLinusN: won't pcm buffers have to be pretty large before we can afford to waste 10ms here and there in dynamically readjusting the cpu clock?
10:31:29LinusNi think the pcm buffer should be fairly large, like 1-2 seconds of audio
10:32:07LinusNmaybe even more, if we want to crossfade
10:32:08preglowif that's the case, there is no problem
10:32:13 Join ze [0] (
10:32:33LinusNhow's the optimizations going?
10:32:46preglowonly problem with buffers that large is that we'll have tons of latency if we change parameters
10:32:50preglowlike volume or eq settings
10:33:22LinusNvolume should be handled by the 1380
10:33:56preglowyes, but what about other settings?
10:34:09LinusNyes, that might be a problem
10:34:47LinusNwe might even have two pcm buffers
10:34:58preglowpeople are going to request things like eq's, and i'm going to write one anyway, heh
10:35:02preglowand it'll be a pain to adjust
10:35:10LinusNone big "raw" buffer, and one smaller for eq stuff
10:35:40preglowbut yes
10:35:47preglowcan you tell me why the jump table isn't working?
10:36:26LinusNmovea.l .Ljumptable, %a4 is wrong
10:36:36LinusNit should be lea.l .Ljumptable, %a4
10:36:39preglowyes, i suspected so
10:36:45preglowthat still gives me an illinstr instruction
10:36:49preglowehh, error
10:36:53LinusNor movea.l #.Ljumptable, %a4
10:37:16preglowlea.l has no immediate form, how can lea.l work?
10:37:35preglowwell, duh
10:37:37LinusNlea.l is Load Effective Address, the equivalent of move.l #xxx
10:37:39preglowit has
10:38:13preglowbut yes, i tried that yesterday, and then it just gives me an illinstr as opposed to just hanging :)
10:40:39dwihnoan improvement! :)
10:40:54LinusNjmp (%a4, %d2.l*4) doesn't do what you think
10:41:12preglowapparently not
10:41:23LinusNit jumps into the table
10:41:38LinusNmove.l (%a4, %d2.l*4),%a4
10:41:45LinusNjmp %a4
10:41:50LinusNjmp (%a4)
10:42:27preglowwell, i think that's queer syntax, then
10:42:55preglowbecause all other instructions would get the contents of that address before acting on it
10:43:05LinusNthe (%a4, %d2.l*4) addressing mode is the same for move and for jmp
10:43:06preglowapart from lea
10:43:33preglowok, then i'm just fundamentally confused
10:43:47LinusNmove reads from the address, jmp jumps to it
10:44:10preglowyes, i figured it did a read, then jumped to that :)
10:44:23LinusNthat's 6502 assembler :-)
10:44:29 Quit ze__ (Read error: 110 (Connection timed out))
10:44:34preglowmight have been tainted by that, then
10:44:48preglowcan't claim i remember it anymore, but oh well
10:45:07LinusNyou should really move the code to iram
10:45:30LinusNand the data
10:45:40preglowi will
10:45:55preglowthis method is moved to iram in rockbox anyway
10:46:06preglows/method/function/, i'm damaged by c++
10:46:30LinusNfunny, i haven't seen "jeq" before
10:46:45LinusNbeq is the normal name for it
10:46:47preglowit's a gas synonym for beq
10:46:58preglowit computes the shortest relativt jump form and substitutes it for that
10:47:11LinusNnice, didn't know that
10:47:18preglownow you do :)
10:47:26preglowall conditional branches have jcc forms
10:47:44preglowincluding bra itself
10:47:55preglowit works nicely now, thanks a lot
10:48:45preglowis aligning before all loop labels a must? i have never seen the align actually trigger
10:48:58preglowi thought all instructions were at least a word wide anyway
10:49:24LinusNthe align isn't neccessary on this target
10:49:40LinusNit is on the sh1
10:49:55preglowgcc uses it, but that wouldn't be the first coldfire issue it's confused about
10:50:45LinusNi think it's standard procedure to align before jump targets
10:51:05preglowyes, but if its not needed, i'll skip it
10:51:26preglowi can't think of a single reason to do it for coldfire, unless i've got a bunch of .bytes before it or something
10:59:22 Quit webguest45 ("CGI:IRC (Ping timeout)")
11:03:05***Saving seen data "./dancer.seen"
11:04:22preglowthere, i don't think this routine gets much tighter than this
11:04:29LinusNlemme see
11:04:42preglowjust testing it first
11:05:20 Join Patr3ck [0] (
11:05:51preglowLinusN: there's something wrong with the lower order cases, but you get the overall idea
11:05:55preglowi'll fix them now
11:07:04LinusNlooks really nice
11:07:23preglowyes, for a first attempt at an asm function, i'm very satisfied myself
11:07:29preglownon-x86 asm function, that is
11:09:25preglowi think i'll try to make the jump table relative, though
11:12:41LinusNpreglow: can you explain your findings about the long compare gcc quirk?
11:13:18preglowwell, i don't know much more than you guys told me yesterday
11:13:38LinusNaha, i thought you were the one that obsevred it
11:13:48preglowi did
11:13:52preglowi might remember wrongly about it putting numbers bigger than 8 in a register as well
11:14:04preglowi'll try to make a test case
11:14:26LinusNfor example, the coldfire can't do this:
11:14:40LinusNcmp.l #12345678,(%a0)
11:15:02LinusNthe destination must be a register
11:15:17preglowknow that
11:15:37LinusNso it has to either move the variable to a register, or the constant
11:16:05LinusNif(myvar == 0x12345678)
11:16:11LinusN res = 1;
11:16:14LinusNtranslates to
11:16:28LinusNmovel #12345678,%d0
11:16:50LinusNcmpl myvar,%d0
11:17:12preglowyes, i see gcc does that
11:17:22LinusNit has no choice
11:17:41preglowof course not, just want to see it doesn't do that if it's got the data in a register
11:19:16LinusNok, i made it put myvar in a register
11:19:32LinusNand now it does cmpl #12345678,%d0
11:20:00LinusNso it works just fine
11:20:07preglowok, then there's not problem
11:20:52LinusNhave you had any problems running at 96MHz?
11:22:19preglowbut ok, now for making the default case
11:22:32preglowsince we really should support higher orders as well
11:24:03LinusNso you have both code and data in iram now?
11:24:19preglowahh, that's what i wanted to ask you about
11:24:35preglowthe .icode definition is in coldfire.c
11:24:41preglowi don't use that file anymore
11:24:44preglowi only have coldfire.s
11:24:55preglowcan i stuff it in the include/private/coldfire.h instead?
11:25:25LinusNno, you should have it in the coldfire.s file
11:25:37preglowthen you'll have to tell me what it should look like
11:26:15LinusNinstead of section .text, use section .icode
11:26:30preglowof course...
11:26:31 Quit jyp (Read error: 110 (Connection timed out))
11:26:36preglowi wish i'd think
11:26:48LinusNme too :-)
11:28:33preglowahh, rockbox makefile only knows how to make .S not .s ?
11:29:08LinusNi suggest you use .S
11:29:27preglowi do
11:29:43preglownow for testing it with actual libflac
11:30:07LinusNshould be coldfire.S
11:30:44preglowyes, done
11:31:05LinusNlunch time
11:39:23preglowhaha, make clean removes coldfire.S all the time :P
11:39:27preglowbut now for a shower
11:40:11 Join cYmen [0] (
11:46:53*ashridah kicks the requirements for an assignment
11:47:05ashridahsodding lecturer has us using gcc -ansi, so, fantastically, no snprintf: (
11:47:08ashridah:( even
11:51:41preglowsnprintf is c99
11:51:59preglowhell, no, it's not
11:51:59ashridahi know
11:52:03ashridahyes it is
11:52:10preglowhell yes, it is!
11:52:14preglowi'm just confused
11:52:24preglowdoesn't -ansi enable c99 stuff?
11:52:26ashridahthe line immediately above it in stdio.h is __BEGIN_NAMESPACE_C99
11:52:34ashridahpreglow: no, it goes WAY back to C90
11:52:55*ashridah mutters
11:53:17preglowwell, the intent is good, he wouldn't want you using some nasty gnu extensions ;)
11:53:50ashridahthat's okay with me. what i'm not okay with is converting numbers into strings the annoying way :)
11:54:12ashridah(and continuously bounds checking the string while i'm at it)
11:55:24ashridahah well, at least i can use strcpy.
11:56:17ashridahhrm. strncpy at least seems to exist for C90
11:58:29preglowyes, strncpy has been around for a long time
12:08:12 Join Ka_ [0] (
12:12:31preglowok, so putting the file name in SOURCES is obviously not enough to get it linked in
12:14:21preglowit complains if i remove the file, but it doesn't assemble it if it's there anyway
12:17:01 Quit [FlaT]Heidel (Read error: 110 (Connection timed out))
12:18:10 Join shw [0] (
12:22:06shwhi! i have one simple question
12:22:45shwis there in plans to make rockbox to other players?
12:23:31preglowapart from the archos players, gmini and iriver h1x0?
12:26:29preglowwhat do you have in mind?
12:26:48shwi bought DivaGem and company will not update firmware for my hardware version and they don't wan't to release firmware source
12:27:22shwso only not original firmware is my last hope to do something with it..\
12:29:18preglowwell, i don't know anything about it
12:29:24preglowwe're quite busy with h1x0 port these days
12:29:38preglowfor there to be a chance, there has to be specifications available for all components
12:30:32 Join [FlaT]Heidel [0] (
12:33:42preglowporting to a new platform is a lot of work
12:34:32shwi imagine that
12:36:04shwbut i'm despered - i didn't know that i'm buying version of a player with the oldest hardware.. a lot of features are disabled
12:36:11shwin one word - sux
12:37:03shwi would even like to help to create that firmware
12:37:49shwbut company don't want to release source of his firmware..
12:37:59preglowwell, are you a programmer?
12:38:31shwc/c ,php,asp a little c# - yes. asm - no
12:41:01preglowwell, i can't promise a thing, but you'll have to find out about the hardware first
12:41:54shwso what i have to know?
12:43:29preglowwell, everything about it, what components are used
12:44:53shwi'll try to do my best :)
12:53:12 Join shx [0] (
12:57:48 Quit DMJC (Read error: 110 (Connection timed out))
12:59:42preglowwhy the hell does coldfire.S end up in OBJS? no wonder it gets deleted
12:59:56LinusNpreglow: i have added support for .S in the makefiles
13:00:02LinusNupdate your cvs
13:00:10preglowahh, great
13:03:06***Saving seen data "./dancer.seen"
13:04:32 Join Chamois [0] (
13:04:35preglowyup, small boost, other things obviously needs looking at in libflac as well
13:04:50preglowit's 76-77% realtime on my test file
13:04:58LinusNin 96MHz? good
13:05:22preglownot much has been optimized apart from that one function
13:05:22LinusNmoving stuff to iram will get you pretty far
13:05:40LinusNhow fast was it before your boost?
13:06:05preglowi'll yank it out and see, my opt has been in cvs for a long time
13:07:17preglowthis only gave a minor boost compared to that, but i think having it in a .S file is cleaner than a c routine that's mostly asm blocks
13:10:14LinusNis the data in iram?
13:10:55preglowdon't think so
13:10:58preglowthis is weird
13:11:15preglowunless i've done something wrong, the stock c routine performs about as good as the asm version
13:11:19preglowand i find that hard to believe
13:11:53preglowit's not even unrolled
13:12:10LinusNunrolled isn't necessarily faster
13:12:34preglowno, that's true, but should be here
13:12:43preglowi even fetch all coefs in advance, and just once
13:12:46preglowc version does not do that
13:12:49LinusNthe cache doesn't like unrolled loops
13:13:25preglowi just have to make sure my routine is what's actually used
13:14:10 Join SeeSchloss [0] (
13:15:43 Join ze__ [0] (
13:15:43preglowall the relevant data is malloced
13:15:52LinusNhow silly
13:16:21 Join Patr3ck_ [0] (
13:18:17preglowif we plan on using just one flac encoder at a time, we could of course make it a static struct
13:18:17 Quit shx ("CGI:IRC (EOF)")
13:18:36LinusNwe should
13:18:49preglowwell, i can't imagine why we'd want more
13:19:01LinusNgah, realloc!
13:19:23preglowhe's not making it easy
13:19:40 Join Aison [0] (
13:24:13LinusNlooks like most mallocs are for metadata
13:24:16 Join DMJC [0] (
13:25:27preglowmy routine is never called, why the hell is that
13:26:17LinusNmaybe coldfire.c is still there?
13:26:56 Quit ze (Read error: 110 (Connection timed out))
13:26:57 Nick ze__ is now known as ze (
13:27:03preglowCONFIG_CPU is declared everywhere?
13:27:36LinusNonly where you include config.h
13:28:58preglowit is defined, i see
13:29:07LinusNand stream_decoder.c doesn't include config.h
13:29:46LinusNuntil after the inclusion of private/coldfire.h
13:29:59preglowi put a #error in the ifdef CONFIG_CPU above it, and it did trigger
13:30:22preglowso it's included somewhere
13:30:37LinusNprobably stdio.h
13:31:05 Quit Patr3ck (Read error: 110 (Connection timed out))
13:33:46preglowjust an unusual place to have it included, but makes sense
13:34:00preglowi wonder why the bloody flaming hell my function is never called
13:34:21preglowi've set it to skip straight to exit, so i should notice
13:34:50preglow.Lexit in my function, it just does rts
13:35:04preglowand stack cleanup
13:35:14 Join webguest47 [0] (
13:35:31LinusNbut are you sure that your file needs an order 8 restore?
13:35:46preglowyes, and i've set my function to handle both cases anyway
13:36:05webguest47do you know this
13:36:07pregloworders higher than eight are only used if you explicitely specify it or use −−best
13:36:14LinusNwebguest47: yes we do
13:36:17webguest47it seems to be an mp3 decoder working with coldfire
13:36:34LinusNthanks anyway
13:37:07webguest47on what are you working today ?
13:37:15webguest47for iriver i speak
13:37:17LinusNcodec optimization
13:37:26webguest47i let you work
13:37:37 Part webguest47
13:37:44LinusNpreglow: !SIMULATOR should be !defined(SIMULATOR)
13:37:57preglowi've removed that as well, but fixing
13:40:41preglowit now gives me an illinstr
13:42:15preglowi copied entire rockbox over again, not just the plug
13:42:34 Quit shw ("CGI:IRC (EOF)")
13:42:38preglowillinstr at start of decode
13:42:42 Join Schee [0] (
13:42:52preglowthat's iram, isn't it?
13:43:03preglowmy function is called!
13:43:12preglowso, the very first instruction is erronous? :)
13:43:16LinusNlooks like the iram code isn't copied to iram...?
13:43:32preglowmost probable
13:43:45preglow rb->memcpy(iramstart, iramcopy, iramend-iramstart);
13:43:47preglowin flac2wav
13:43:51preglowthat should do the trick, no?
13:44:15LinusNit should be done in the flac2wav plugin
13:44:21LinusNand it is, in cvs
13:44:33preglowyes, it is here as well
13:45:47preglowi'm doing nothing special to use iram apart from .section .icode
13:47:20 Quit Ka_ (Read error: 104 (Connection reset by peer))
13:48:30 Join Ka_ [0] (
13:49:06LinusNpreglow: i think i see the problem
13:49:17preglowLinusN: goodie
13:49:39preglowanything nasty?
13:50:06LinusNnot sure
13:50:27LinusNit might be that the idata copy isn't properly included when creating the .rock
13:50:55preglowoh? it worked before, with coldfire.c
13:51:16preglowwhich was also marked .icode
13:55:24 Quit ashridah ("Leaving")
13:59:21LinusNno, it is included
13:59:50LinusNso either it isn't properly copied, or something else is wrong
14:00:51 Quit lostlogic ("Going to the moon")
14:01:45 Quit SeeSchloss (Read error: 110 (Connection timed out))
14:03:12preglowwell, 1001000 should be the very first instruction, which is a lea
14:03:23LinusNpreglow: you must reconfigure
14:03:24preglowwhich is perfectly good and works fine from sdram
14:04:00preglowLinusN: actually, i'm pretty certain i just did, but will try again
14:04:56preglowyes, still crashes
14:05:51LinusNi guess you have tried make clean?
14:06:22preglowyes, every time i do a change
14:10:26HClwhat were the keys to get the iriver to 96mhz?
14:10:38HCldebug, and then... ?? and.. up.. ?
14:11:13LinusNdebug->view i/o ports
14:12:02preglowmaybe the routine didn't make it to iram the last time?
14:12:24HClholy shit.
14:12:30HClrockboy is actually pretty playable
14:12:37HClmario works like, pretty good o.o
14:13:27DMJCwhat do you need to do to get roms loading?
14:13:57 Join methangas [0] (
14:14:00HCli died
14:14:16HClbut i got to the 6th mario block
14:14:19HCland several enemies
14:14:21pabsHCl: playable eh?
14:14:28pabsHCl: what system?
14:14:49DMJCwhat do you need to do to make it work?
14:14:55HClseems i made a bug on the shutdown thing
14:14:57DMJCeg what files do you put on the iriver and where
14:15:15HClits broken on shut down, i think.
14:17:31HCldoh, stupid me.
14:17:44HCltrying to play pokemon but not setting it to 96 mhz first
14:18:43preglowLinusN: well, i sure as hell don't get it, and my debugging facilities are somewhat limited
14:18:47DMJCok so I have the .zip file on my iriver now
14:18:53 Join ze__ [0] (
14:18:54DMJCso how do i get the game to run?
14:19:06preglowso i think i'll just go the the post office and pick up a book instead of banging my head against this
14:19:37HClyou need to unpack the zip onto your iriver, reboot, place some gameboy games on it
14:20:25DMJCwhere do the roms need to go?
14:20:38LinusNpreglow: wait a sec
14:20:41DMJCI have on there already
14:20:57preglowLinusN: can do
14:21:37HClhello there welcome to the world of pokemon blahblahblah
14:21:41HClthis works absolutely fine
14:21:51DMJCI can't make the rom load
14:22:08HClshouldn't have much trouble getting it running at full speed if we manage to get 140mhz working - rockboy does not require hdd access during emulation
14:23:26preglowHCl: not to mention utilizing sram
14:23:28HCli need to implement that managing of the 16 lost scanlines asap
14:23:39HClpreglow: yea, moving the interpreter core to sram might be enough
14:23:43preglowHCl: you need a skip top 8 and botton 8 core as well
14:23:59HCli'm planning on adding several
14:24:08HClskip 16 top/bottom selectable by hold
14:24:12preglowa panning mechanism would be great
14:24:19preglowso you can select which you'd like to skip
14:24:21HCldistribute the 16 skipped over the screen evenly
14:24:30HClskip the top 8 and the bottom 8
14:24:48preglowthe first will look like shit
14:24:53DMJCI can't get rockboy running
14:24:54*HCl continues playing pokemon
14:24:59DMJCIncompatible version
14:25:08HClyou need to unpack the entire .zip
14:25:14DMJCI did
14:25:43HClits working way better than what i had expected
14:25:59*HCl goes to get dressed and fix those few bugs
14:28:15preglowHCl: indeed, it works a lot better than i would have thought
14:30:01 Quit ze (Read error: 110 (Connection timed out))
14:30:01 Nick ze__ is now known as ze (
14:30:15DMJCuh rockboy isn't in there
14:30:26preglowDMJC: it is
14:30:35preglowDMJC: it's in viewers/
14:30:54DMJCso how do I use it?
14:31:00DMJCplace a rom in a sub folder?
14:32:39preglowDMJC: place it wherever you like
14:32:54preglowDMJC: just make sure you've got rockbox set to display all file types
14:33:04preglowthat last might not even be necessary, i don't remember
14:33:15DMJCwhere is that setting?
14:33:20*HCl goes to work on the few remaining bugs
14:33:22preglowgeneral settings
14:34:41LinusNpreglow: the iram section is never copied
14:35:12LinusNor rather, the iram contains 0's
14:35:14preglowLinusN: that surely explains it, why is it not?
14:35:37preglowa plugin format that contained section identifiers would be sweet
14:36:09preglowor at least separate sections with different load addresses
14:36:14preglowbut that would be pretty device specific
14:36:31DMJChow do I start the game?
14:36:45preglowDMJC: just select it
14:36:58DMJCI already got the rom loaded
14:37:06preglowwell, just try buttons
14:37:08preglowrecord is start
14:37:11DMJCah k
14:37:32HCli must admit i don't really see why it crashes on quit at the moment
14:42:42preglowLinusN: but i've got to go pick up this package before they close, i'll be back in around half an hour
14:45:30 Quit DMJC ("Leaving")
14:50:11 Join shx [0] (
14:54:35LinusNpreglow: .section .icode,"ax",@progbits
14:55:47HClLinusN: how big is the iram and how would i go about moving rockboy's cpu core there?
14:56:30LinusNthe iram space available for plugins is 32kbytes
14:57:18LinusNlook in flac2wav.c to see how to copy the idata/icode sections to iram
14:57:37LinusNthen you declare the functions you want in icode like this:
14:58:11LinusNvoid myfunc(void) __attribute__ ((section(".icode)));
14:58:18LinusNvoid myfunc(void)
14:58:35HCland how can i check how big my function actually is in the end..?
14:59:55 Join lolo-laptop [0] (~lostlogic@
15:00:00 Join edx__ [0] (
15:02:26 Join R3nTiL [0] (~zorroz@
15:03:07***Saving seen data "./dancer.seen"
15:03:46 Quit Ka_ (Read error: 110 (Connection timed out))
15:05:11 Join Ka_ [0] (
15:06:20HClgreat, seems to fit
15:08:41 Join ripnetuk [0] (
15:09:18ripnetukby the way, i made a simple plugin for iriver that jumps to the original firmware. useful when it boots to rockbox by accident, and i wanted iriver...
15:09:30ripnetukis anyone else interested in this?
15:09:54pabsthat's kinda cool
15:09:56LinusNisn't it easier to just reboot?
15:10:04ripnetukcant decide :)
15:10:38pabsit's more keys to use the plugin, i think LinusN is right
15:10:48pabsbut it's still cool :D
15:11:01ripnetuki guess...
15:11:21 Quit edx (Read error: 110 (Connection timed out))
15:11:24LinusNwell, perhaps if you name the plugin ___.rock and put it in the root
15:11:52LinusNthen you just play it to reboot
15:11:53 Nick edx__ is now known as edx (
15:11:55ripnetuki think its easaier with one hand, as you dont have to have record down... anyway, not important - i use it, and thats why GPL is great
15:13:41ripnetukthanks for the idea of ____.rock in root - will make it even easier :)
15:13:54*HCl wonders what kind of speed increase iram adds..
15:14:11ripnetuki will let u get on with it :) nice work on speedup btw - IS now fast enough to impress me
15:14:30 Quit ripnetuk ("Leaving")
15:15:58LinusNHCl: iram gives a tremendous boost
15:18:21preglowlets see what we can see
15:18:40preglowLinusN: works excellent
15:18:44 Join einhirn_ [0] (
15:18:47preglowLinusN: and quite a boost as well
15:18:52 Quit einhirn_ (Client Quit)
15:19:06preglowLinusN: what does the extra information in the .section line do?
15:19:36preglowLinusN: went from 75% realtime to about 83%
15:20:44HClokay, shutting down works fine, its hard to tell at what speed its running, i added some optimizations..
15:21:01 Quit Sando (Read error: 113 (No route to host))
15:22:27HClit also speeds up on its own now, rather than having to use the debug function... i added cpu_boost to the plugin api..
15:22:50preglowcpu_boost is in the plugin api
15:23:04HClnot last time i checked..
15:23:07preglowignore me
15:23:15preglowi could have sworn i saw it added
15:23:43LinusNpreglow: it tells the linker what the section contains
15:23:58pabssay, where is rockboy, anyway? i'd like to try it
15:24:16HClpabs: hold on while i add another cpu-intensive function to iram
15:24:25pabsHCl: k
15:25:04pabsit's times like these i love autologging
15:25:17pabssince i'm going to go back and reread LinusN's optimization comments later
15:25:17HCli'm really eager to add markun's greyscale patch to it
15:26:03pabsthat should be a speedup, no?
15:26:24HClit'll look better, at least
15:26:57pabswasn't the gb 2 bit as well? it seems it'd be less mucking around with the graphics if it's a straight map, instead of dithering
15:27:01HClyes, it was
15:27:39pabsthen again, i guess a 2-bit to 1-bit translation is negligible in terms of cpu time, also the graphics will be larger
15:27:40HClwell, at the moment its just checking 0x2 against the gameboy format, most significant bit of the graphics, and sets it to black if 1 and white if 0
15:27:42pabsso hmm
15:28:01pabstough call then
15:28:03HClwith 2 bits, it'll be checked against 0x3, and rather than setting
15:28:06HClit has to bitshift it
15:28:15HCland or
15:30:29*HCl wants a new episode of bleach :/
15:30:40HClon my ftp
15:30:50pabsah okay
15:31:04HCli forgot to add the lcd function into iram, i'll fix that in a bit
15:33:44HClwo fucking ha.
15:33:47pabscan i just copy the .rocks traight over?
15:33:56HClits almost running at full speed
15:34:03HClit pretty much went through to the mario demo immediately
15:35:28LinusNHCl: you have your own lcd frame buffer?
15:35:36HClpretty much, yes.
15:35:42LinusNis that in iram?
15:35:42HClthough i get them one scanline at a time
15:35:45HClnot yet.
15:35:57LinusNthat should give a nice boost as well
15:36:08HClits almost running at full speed already
15:37:19HClbut first i need to rewrite the button driver
15:37:30HClwell, actually, i'll do lcd first >.>
15:37:37HClthen the button driver
15:39:21HClhow do i put certain variables in iram?
15:39:33preglowHCl: yes, get rid of the button queue
15:39:44LinusNint myvar __attribute__ ((section(".idata")));
15:40:03preglowcouldn't we make a define for that?
15:40:10preglowit might be applicable to other platforms as well
15:40:38LinusNyou mean for platforms without iram?
15:40:48preglowLinusN: more like platforms with iram
15:40:56preglowLinusN: for platforms without, it'll just be an empty define
15:41:00LinusNthe syntax is exactly the same for all platforms
15:41:03preglowLinusN: ok
15:41:13preglowLinusN: how does the simulator handle it? idata = data?
15:41:38LinusNthe __attribute__ is inside #ifndef SIMULATOR
15:42:01LinusNbut i agree that a #define might be more general
15:42:11preglowwell, yes, and you'd avoid having to check for simulator all the time
15:42:21HCli have to agree with preglow here
15:42:25preglowit can be an empty define for simulators as well
15:43:03HCli went over the top, doesn't fit in iram :/
15:43:21preglowi wonder how we'll handle stuff contesting over the use of iram
15:43:57LinusNwe'll give a fixed space to each codec
15:44:08LinusNthen a little to the plugins
15:44:12HCli have about 4kb left
15:44:17preglowLinusN: yes, but we'll have to add relocation tables to the plugin format, no?
15:44:39preglowno? isn't iram addresses hard coded?
15:44:51LinusNthe plugins will have a dedicated space, just like the codecs
15:45:17LinusNall plugins will use the same addresses in iram
15:45:35preglowahh, ok, so we can never have to concurrent codecs running?
15:45:43preglowi think zagor might complain
15:45:59LinusNwe'll give a fixed space to each codec
15:46:09LinusNonly 2 simultaneous codecs
15:46:19LinusNand it's zagor's idea :-)
15:46:36preglowand how will we handle the fact that a codec might run in either slot? the addresses will have to be fixed up somehow
15:47:04preglowthe plugin format does not allow for this right now, as far as i know
15:47:13preglowwe'll have to add relocation tables
15:47:22preglowso i propose that if that is to be done, we also add different sections
15:47:22LinusNfor the codecs, yes... :-(
15:47:30preglowso we don't have to manually copy to sram
15:47:33LinusNdifferent sections? why?
15:47:42pabsHCl: "least significant bit", except for, you know, the ground :(
15:48:09preglowLinusN: well, so code belonging in iram can be put in iram by the loader, not by a memcpy in the plugin as is now
15:48:23pabsHCl: almost full speed though!
15:48:28LinusNis that really that much of a problem?
15:48:32preglowLinusN: no
15:48:53LinusNbut the iram code might have to be PIC
15:49:18HCldarnit xD
15:49:24HCli keep dying in this hole
15:49:26HClthat i can't see
15:49:30HClbecause it falls off the screen
15:49:41HCltime to work on the scanline thing and the new button driver
15:49:58pabsHCl: yeah that's what i was complaining about
15:50:03pabsHCl: i'm gonna try another game :D
15:50:14 Quit R3nTiL ()
15:50:40HCl :p
15:51:17preglowLinusN: pic code is position independent at the assembly level, or?
15:51:50preglowrelocation tables and fixing up the addresses that need it would be fastest
15:52:02preglowit'll probably require significant objcopy magic
15:53:10preglowbut it's not impossible, i think nasm has an example for a simple object format like this
15:53:35LinusNanother approach is to have two instances of each codec
15:53:47LinusNlinked at different addresses
15:54:00preglowahh, yes, a very much simpler one at that
15:54:17preglowplugin files will be larger, but that's not much of an issue
15:54:33preglowi think that would be a fine starting point
15:54:52*pabs tries out gargoyle's quest
15:55:06preglowso i vote for it
15:56:05pabsHCl: haha final fantasy adventure runs
15:59:49HCl :)
15:59:56HClhow's the speed?
16:00:00*HCl works on the new button driver
16:00:21pabshold on i'm fighting the first monster
16:01:12pabsHCl: about the same as mario, a wee bit slow but totally playable
16:01:25pabs"poop was a figher of the empire Glaive"
16:05:56pabsi have to disappear for work rela soon
16:06:01pabsHCl: you have this in a cvs repo somewhere?
16:08:34pabsHCl: btw it'd be cool if play brought up a menu that let you save a dump, a screenshot, or quit
16:08:44HCl :P
16:08:45pabsHCl: or load a dump
16:08:51HClyea, i'm working on that
16:08:53pabsHCl: k
16:08:56HClits on my todolist
16:09:05HCland you're right, i need to check in my local cvs
16:09:07pabsHCl: final fantasy adventure and final fantasy legend 3 both work fine too
16:09:28pabsall about the same speed
16:09:36pabslike, a little slow, but totally playable
16:09:49preglowbut ok, i'll start cleaning things up so it can be put in cvs sooner or later
16:10:37 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
16:12:12pabs"Sprite used dive again'st ASS"
16:12:38pabs"CRAP used Battle against Diviner."
16:12:51HCl :P
16:12:55pabsthis is great
16:13:14pabs"Diviner used Kick against ASS"
16:13:28HClok ok.. hm.
16:13:43 Join DeadMan [0] (
16:14:07DeadManGood day to thee sexy devs and all. How are things proceeding? :)
16:14:23LinusNfine, i guess
16:14:24*HCl thinks he could waste another 144 bytes for a proper scanline dropping mechanism
16:14:33HClpretty good, rockboy is running at near full speed
16:14:55DeadManDost thou haveth realtime MP3 playback aseth yet....eth?
16:15:24preglowpay me and i'll see to working on it full time :]
16:15:35pabs"ASS used Long against Diviner. Diviner fell."
16:15:51DeadManare we stuck at 96Mhz on Coldfire?
16:15:57preglowstuck, no
16:15:59LinusNnot stuck
16:16:00preglowbut it's the current maximum
16:16:20DeadManmethinks iRiver desingeth poorly :)
16:16:22pabshey i remmeber when i would have killed for 96MHz
16:16:39pabsall i had was a 24MHz box and i was like "if i had 100 i could do EVERYTHING"
16:16:58pabs"14 EXP up! Received 560GP"
16:17:15DeadManhow much processing does decoding a libmad normally take in comparison to the Coldfire at 96Mhz then?
16:17:35LinusNDeadMan: i don't understand the question
16:17:40DeadManNo do I
16:17:58preglowthat's some funky sentence construction you've got there
16:18:03DeadManI dunno was trying to compare beans to grapes I guess :)
16:18:19preglowit runs at around 50% realtime now
16:18:38DeadMans-t-u-t-t-e-r ;)
16:18:51LinusNpreglow: i'd like you to commit that
16:18:55DeadManOgg will take major work then
16:19:04preglowLinusN: coldfire.s?
16:19:05*HCl hrmz.
16:19:33LinusNpreglow: don't you have some libmad stuff uncommitted?
16:19:41preglowLinusN: i do, i'd like to clean it up first
16:19:49preglowLinusN: can have a look at commiting it tomorrow
16:19:59DeadManI tell ya. If I get just MP3 decoding with gapless I'd be happy. Everything else is just extras to me.
16:20:07preglowi've got bits and pieces uncommited everywhere
16:20:20DeadManoh and use of remote of course
16:20:30DeadManwhich I take it has taken a backseat for now?
16:20:48preglowanyone is free to work on whatever he/she wants
16:20:53preglowit's not my focus, that's for sure
16:21:27DeadManthat's ok. can live without it for a little while I guess
16:22:15DeadManI'm going to make a PIZZA!
16:22:33LinusNyou're about to throw up?
16:22:53DeadManI left my freezer door ajar for an hour os so....good job I noticed it had just began to thaw out :P
16:22:59DeadManor so
16:23:12DeadMansilly me
16:24:07pabsHCl: burger time works too
16:24:16pabsHCl: except i thought a monster was a bonus and died :(
16:24:21preglowLinusN: you think having the jump table use relative addresses would change anything at all?
16:24:29preglowLinusN: 'cause that seems to be what gcc is doing
16:25:05preglowi'll save that for when i'm bored
16:25:08LinusNyes i do
16:25:43preglowLinusN: do you know what a .swbeg directive does?
16:26:01preglowit's not documented anywhere i see
16:27:34HClpokemon looks ugly in black and white
16:27:51LinusNpreglow: it's the beginning of a switch table
16:28:02preglowLinusN: i gathered as much, but does it have any function?
16:29:02preglowor is it just some kind of syntactic sugaring?
16:29:22HCllet me fetch my cam
16:29:25HCland make a pic of pokemon running
16:30:55LinusNi have no idea
16:31:47DeadManOne of the great mysteries of life
16:31:48pabsHCl: woo i made three burgers
16:32:19 Quit Patr3ck_ ()
16:32:29DeadManIf you find out what a .swbeg does you become one of the chosen few and shall ascend to heaven
16:32:30pabs(burger time)
16:33:13LinusNpreglow: seems to be ancient compatibility stuff
16:33:18HClbut first
16:33:22HCli'm gonna add grayscale
16:33:27HClwith markun's patch
16:33:34pabsHCl: haha tetris is playable too
16:33:40pabsthe scanlines are really an issue
16:33:40DeadMangreyscale on iRiver? w00t!
16:34:49LinusNThe .swbeg <N> is an assembler directive that causes the switch table
16:34:50LinusN size to be inserted into the object code so that disassemblers, for
16:34:50LinusN example, can identify that it is the start of a switch table.
16:36:01DeadManI was about to say that! ;)
16:36:16DeadManwell I was googling and I saw mention of tables and gnu that's about it lol
16:37:21HClpabs: yea, i'm gonna fix that in a bit...
16:39:11preglowLinusN: great, think i'll just ignore it, then
16:39:24LinusNdo so
16:42:59LinusNpreglow: the relative table may not buy you anuthing
16:43:34LinusNthe iram is 32-bit, so if the table is aligned, the performance will be the same
16:43:51 Join stripwax [0] (
16:45:23LinusNtime for me to go
16:45:36LinusNcu around
16:45:46 Part LinusN
16:46:36 Part stripwax
16:51:57 Join mecraw [0] (~mecraw@
17:03:10***Saving seen data "./dancer.seen"
17:05:31preglowHCl: i can't get zelda to work here
17:06:00preglowi just can't progress from the start screen
17:22:06 Join condor9 [0] (
17:24:01 Quit condor9 (Client Quit)
17:25:07 Join webguest24 [0] (
17:25:53 Quit webguest24 (Client Quit)
17:30:28 Quit Nibbler ("blubber")
17:38:46 Join Nibbler [0] (
17:38:54 Quit Nibbler (Read error: 104 (Connection reset by peer))
17:41:34 Join Nibbler [0] (
18:09:40 Join stripwax [0] (
18:10:57 Join rasher [0] (
18:33:54 Nick kergoth`zzz is now known as kergoth (
18:37:18 Quit preglow (" :-) :-) :-) :-) :-) in peace....")
18:38:33 Nick Lynx_ is now known as Lynx_awy (HydraIRC@
18:45:17 Quit mecraw (Read error: 54 (Connection reset by peer))
18:45:27 Quit stripwax ("Trillian (")
18:46:05 Join Stryke` [0] (~Chairman8@resnet-241-86.resnet.UMBC.EDU)
18:53:20 Join markun [0] (
18:55:22HClgrayscale rockboy
18:57:23markuncongrats :)
18:57:33 Join AC [0] (
18:57:43markunAC: helo
18:58:41ACmarkun: whats the state of grayscale?
18:59:28markunIt's working, but I have to cleanup the code a bit. HCl just made grayscale rockbox with my grayscale patch. He's taking a picture of it now.
19:01:45ACsounds fine :)
19:02:16markunDrawing 1-bit bitmaps is very slow right now, so drawing a full screen of text is a lot slower.
19:02:43HClgee, pokemon has a sort of demo
19:03:13***Saving seen data "./dancer.seen"
19:03:16*AC watches now simpsons..
19:03:40HClyay for grayscale animation
19:04:25markunLooks good? Is it slower?
19:05:30HCldoesn't seem much slower
19:06:16HClat least pikachu has some shading now
19:06:27HClits on my ftp, and you need to update rockboy to the rockboy.rock on there
19:07:30markunok, I'll try it.
19:07:35HCli need to charge my camera's batteries a little
19:11:06HClit has my new button driver too
19:13:13markunIt looks very nice. It's the first time I tried rockboy.
19:13:33HClwhich rom are you trying?
19:14:00 Join R3nTiL [0] (~zorroz@
19:14:03HCli got the same one running at the moment
19:17:06 Join thegeek [0] (
19:18:14 Join nozomiyume [0] (
19:19:56rasheroh my
19:20:06rasherrockboy is 50% height
19:20:18HClyou need to get the rockboy.rock
19:20:21HClon the same ftp
19:20:36HClit needs the rockbox.iriver from the zip
19:20:40HCland the rockboy.rock in the ftp
19:20:59HClmore good news
19:21:00CoCoLUSpictures :)
19:21:00HClsavegames work
19:21:13HClstill charging my battery
19:21:16HClhold on, let me try
19:21:29CoCoLUSspeed is unchanged?
19:21:52HClcompared to what?
19:21:57HClits near-fullspeed
19:22:01markunWonder if sound will be hard.
19:22:15CoCoLUSso the dynarec is working?
19:22:48HClwithout dynarec
19:23:19HCli made two pictures
19:23:21CoCoLUSi thought it's incredible slow? you found some optimizations? :)
19:23:25HCllet me see if they're ok
19:23:31HClCoCoLUS: no, this is just the 96mhz
19:23:34HClinstead of 11
19:24:54markunrasher: Does it work for you now? I don't have any problems.
19:25:53rasherstill half height.. not sure if I failed to overwrite the rockboy.rock
19:26:02HCllet me upload my pics
19:27:22rasherrockbox stopped on shutting down
19:29:40rashermaybe because usb was connected but not detected
19:29:44rasheror something
19:29:49 Quit thegeek (Read error: 104 (Connection reset by peer))
19:31:11rasherah, there we go
19:31:19rasherI had indeed misplaced the rockboy.rock
19:33:12rasherrunning fine
19:33:18rasherbut I don't know how to control it
19:33:25HClmade a pic
19:36:21HClmy pics are crappy
19:36:22HClbut yea.
19:36:25HClits a pic :x
19:37:10CoCoLUSwhere? :)
19:38:11CoCoLUSlooks nice
19:38:35CoCoLUSjust a quick guess please, what do you think, is the iriver fast enough to emulate the gba?
19:38:41HClno way.
19:38:53HClif only for gba being 320x288
19:39:08CoCoLUSwell thats downscaleable :)
19:39:20ACHC1: looks quite impresive :)
19:39:22markunWhat about rockbox for gba? :)
19:39:57ACcan't wait until rockyboy gets into cvs
19:40:20pabsHCl: gargoyle's quest doesn't work... you suck!
19:40:22pabsHCl: :-D
19:40:26HCl :(
19:40:28HClwhats not working about it?
19:40:35pabsHCl: not sure exactly, migt be a bad rom
19:40:46pabsHCl: it loads and then appears to die
19:40:47HClpreglow said zelda wasn't working
19:40:50HCllet me check that out..
19:41:06HClzelda won't be playable with no diagonals anyways
19:41:40pabsHCl: either way everything else i've tried so far works great.. here's a list if you want to post it somewhere: tetris, final fantasy adventure, final fantasy legends 3, and burger time
19:41:48HClits okay
19:41:55HClit should have the same compatibility as gnuboy
19:42:06pabstetris is kind of impossible though
19:42:12pabssince you can't see the bottom two rows
19:42:15HCllol. i'll fix that.
19:42:16 Join Renko [0] (
19:42:22pabsyou better!
19:42:32pabsi need my tetris
19:43:02pabsHCl: :D
19:43:09 Quit markun ()
19:43:13pabswonder what rockbox needs that i can write
19:43:17pabscalendar, document viewer?
19:43:30CoCoLUSthe rockbox.iriver from the on your ftp?
19:43:34pabs(other tan plain tex)
19:44:03HCl :X
19:44:06HCldarnit :x
19:44:12CoCoLUSits the same here :P
19:44:19HClCoCoLUS:, and update rockboy.rock from the ftp
19:44:37CoCoLUSand rebooting for this damn setup
19:46:50 Quit R3nTiL ()
19:47:07CoCoLUSit runs
19:47:23CoCoLUSis it normal that the main button quits rockboy?
19:47:24HCli don't understand what preglow says
19:47:26HClit works fine
19:47:30HClyesssss :X
19:47:37pabsi'm gonna have to try zelda
19:47:39HCli ran out of buttons and it was the only button left
19:47:43HClzelda works fine
19:47:44HClnice intro
19:47:56pabsHCl: i think A->B quits msot other apps
19:48:01pabsHCl: might want to switch it to that
19:48:21pabsHCl: when the remote buttons are finished will you be able to distinguish those from others?
19:48:27CoCoLUSi maccidently pressing it all the time
19:48:27HClthey are finished
19:48:29rasherno, stop quits most apps
19:48:29CoCoLUSkind of sucks :)
19:48:33HCland yes
19:48:48pabsHCl: k, using stop and play on the remote for a and b would be awesome
19:49:09pabsHCl: that'd give you basically the old gameboy
19:49:28rasheralso, the jpeg viewer is in recorder res :(
19:49:37pabsrasher: most of the .rocks i've tried exit on AB, not on stop
19:49:50rasherpabs: which ones?
19:50:05pabsrasher: lemme check
19:50:37pabsrasher: the problem here is that i'm on crack
19:50:39pabsrasher: :)
19:50:45CoCoLUSdoes rockboy change the cpu freq. itself?
19:50:49rasherAh :)
19:50:57HClCoCoLUS: yup.
19:51:19pabsrasher: i think i subconsiously avoid AB since it always fucks me in the iriver firmware (turns on different play modes), so i think of stop as the bottom button
19:51:39CoCoLUSnot to sound disappointed, but super mario still seems kind of slow
19:52:04pabsCoCoLUS: given the time that's been invested so far, i think it's pretyt damn good
19:52:12CoCoLUSyeah it wasn't a complaint
19:52:14pabsi bet he'll get it full speed
19:52:27pabsHCl: work harder dammit
19:52:42pabs(i'm on post-lunch sugar high)
19:53:24CoCoLUSbut a few bottom lines seem to be cut, yes :)
19:53:25rasheroh boy, the jpeg viewer isn't restricted.. I wonder what was happening
19:53:42rasherI'm on my iRiver!
19:53:59rasher(a picture of me, using the jpeg viewer)
19:54:01HCljpg is adapted for 4bit grayscale
19:54:17HClnot 4
19:54:32rasherWell the size was somehow like 1/4 of the screen
19:55:07rashernow if only I had a decent camera
19:57:29CoCoLUShm.. now where do i find a tetris rom...
19:58:36Chamoisi have one if you want
19:59:11CoCoLUSyeah plz
19:59:25Chamoisis it possible by irc ?
19:59:36CoCoLUSs try :)
20:02:45 Join geoff_o [0] (
20:04:38CoCoLUShow much work would it be to let the user configurate the button layout?
20:04:45CoCoLUSsimple text file with mappings?
20:05:21HCli dunno, anyone can change it and recompile in the source
20:07:18CoCoLUSyeah thats the most user friendly and easiest variant, i'm sure ^^
20:07:44CoCoLUSok so tetris kinda unplayable :(
20:07:50ACit shouldn't be that hard to make a user configuration
20:08:12HCli'll start adding lcd modes..
20:08:42ACIf the rockyboy source is in cvs, i will try to make it
20:10:41ACbut next step for me is to optimize libwavpack
20:11:20 Join thegeek [0] (
20:12:07 Join frank_ [0] (
20:12:40pabsAC: talk to HCl
20:12:43pabsi could use it there also
20:12:48CoCoLUSyeah... having perfect gb emulation but no music playback seems funny :)
20:13:03pabssince i want to make some modifications to rockbox, but i can't do that and keep rockboy around
20:13:16 Quit Renko ("cheerio")
20:13:54*AC switches to linux
20:13:57 Quit AC ("CGI:IRC")
20:14:44ChamoisHCl : rocboy is is not any more in fullscreen?
20:15:01CoCoLUSit is actually in fullscreen +1 :P
20:16:07Chamoiswhy on my iriver it is in 1/4 scrren ?
20:18:24Chamoisanyone ?
20:18:54[FlaT]Heideltried the new rockboy.rock from HCl's ftp?
20:19:44[FlaT]Heideltry again :p
20:19:58[FlaT]Heidelearlier today that worked for somebody
20:20:20Chamoisi speak about the new version with grayscale
20:21:28[FlaT]Heidelthen i can't help you. i'm just an innocent bystander ;)
20:23:06CoCoLUSnothing to see here :)
20:23:36Chamoiscocolus : did you try tetris with teh grayscale version from HCL ftp ?
20:24:37rasherChamois: you need to also download rockboy.rock and put that in .rockbox/viewers/ is not enough
20:24:56Chamoisi did
20:25:00Chamoismaybe i am silly
20:25:11 Join AC [0] (
20:25:40rasherwell, worked for me[tm]
20:26:17rasheroh so, the jpeg viewer only does resizing in entire pixels?
20:26:31Chamoiswork for me now
20:26:36Chamoisi am very stupid
20:26:37ACi want to move a function to iram
20:26:38AClong unpack_samples (WavpackContext *wpc, long *buffer, ulong sample_count) __attribute__ ((section (".icode")));
20:26:40CoCoLUSthose different display modes, will they be auto-selected or does the user have to do this?
20:26:47ACwhats worong with it?
20:27:12ACdo i need to include a special file?
20:28:48rickst131ok, this may be a dumb queston, but i just installed rockbox the other day...
20:29:00rickst131how do i get to things under viewers
20:29:10rickst131like rockboy, jpeg, etc
20:29:22ACwait until they are in cvs?
20:29:24 Quit thegeek (Read error: 54 (Connection reset by peer))
20:29:59Chamoison the HCL ftp
20:30:12rickst131i mean once im in rockbox.
20:30:17rickst131how do i navigate to them.
20:30:21ACrockboy and jpeg viewer are not in the current cvs.. so use HC1's version or wait
20:30:34rickst131i am using HC1's version.
20:31:20rickst131when i load rockbox up on my iriver, i know how to get to plugins, but i dont know how to get to the viewrs
20:32:36Chamoisonly to select a rom and click on the navi button
20:33:03rickst131oh, ok... that makes since... thanks.
20:33:07rickst131i see how they work now.
20:33:14rickst131thanks a lot.
20:36:11rashercan't you move around in the jpeg viewer when viewing full size?
20:36:55rasheror is this just a matter of lacking keybindings?
20:37:32Chamoislack i think
20:40:01 Join Sucka [0] (
20:40:13*AC watches some tv now
20:47:21rasherhrm, it seems like sometimes the jpeg viewer will end up leaving the screen using the wrong colours
20:49:29rickst131same problem here
20:49:49rickst131im getting a lighter version of rockbox sometimes.
20:52:37 Join einhirn [0] (
20:53:22HClAC: you need to copy iram like in flac2wav.c
20:58:08 Part geoff_o ("Kopete 0.9.1 :")
21:03:14***Saving seen data "./dancer.seen"
21:06:57ACHC1: will try it later at some functions, which are called very often
21:14:25 Quit xen` (Read error: 104 (Connection reset by peer))
21:14:43HClhcL!!! L L L L L ;.;
21:14:46HClmrf :(
21:15:06CoCoLUShcell? did you see my question about the display modes above? :)
21:15:23HClthe user,.
21:16:15CoCoLUSanother key needed :P
21:16:34HClthats already there, the hold button
21:16:49CoCoLUSare there just 2 display modes?
21:17:05HCli have several planned. but for now the hold button will do.
21:17:22CoCoLUSwell please just make tetris your priority :P
21:24:26HClbit busy, but let me try to get that sorted
21:39:11HClok, its not working, and i don't want to debug it at the moment
21:43:24pabsi just put a tom of roms on my iriver
21:43:25 Quit Chamois (Read error: 54 (Connection reset by peer))
21:43:26pabsthis is great
21:44:00pabswonder if metroid 2 will work
21:44:13pabsholy crap
21:45:17 Join Chamois [0] (
21:48:10 Join preglow [0] (
21:48:31HClpreglow, found no trouble with zelda..
21:48:41preglowHCl: it hangs on the title screen here
21:48:45preglowperhaps i've got a glitchy rom
21:49:00preglowi took my player with me to watch a movie right now
21:49:06preglowtried playing mario, it locked up :D
21:49:11preglowwith backlight on and all
21:49:29HCldid you have the grayscale version..?
21:49:30preglowi had to twist a piece of chocolate wrapper real thin to reset it
21:49:52HCli used markun's patch to get 2bit grayscale to work on my ftp with rockboy.rock on my ftp
21:50:24HCland now i'm tired and not gonna work on it anymore today
21:51:20preglowi've got a ton of patches to make
21:52:01ChamoisHCl : will you commit the gray pathc into CVS ?
21:55:01 Join jyp [0] (
21:55:25HClChamois: thats something for markun, really
21:55:37HClas far as i know amiconn wanted a different format
21:56:25HClpersonally, i prefer markuns approach and i'm going to have to access the lcd directly if what amiconn wants will be in cvs
21:57:36preglowwhat approach do each of them want?
21:58:02HCl8 bits per pixel format, 0-255
21:59:26 Join midk [0] (
22:00:03preglowwho wants that?
22:00:03HClmarkuns is just the native format of the lcd
22:00:11HCl4 pixels per byte
22:00:20preglowhaving native format is really going to complicate stuff
22:00:28preglowthere'd have to be both
22:00:36preglowan option to use it directly might be nice
22:00:37HClpretty much
22:00:41preglowbut i'm all for amiconns approach
22:00:53preglowplugins that don't rely heavily on speed should use that
22:01:01 Join Camilo [0] (
22:01:04HClbut rockboy does..
22:06:25ACpreglow: can you give me some starting points to optimize libwacpack?
22:07:35preglowAC: i have no idea how wavpack works internally, so unless you have profiled it, no, i can't
22:08:08ACpreglow: i have profiled it
22:09:02preglowAC: do you have the info available somewhere?
22:09:07pabsnow i'm playing casino funpack
22:09:44preglowAC: gimme ten minutes to eat and i'll have a look
22:10:08ACpreglow: i have profiled it with vc under windows, but i should give enougth infos to see what functions we need to optimize and how
22:11:15 Join amiconn [0] (
22:11:57 Join Bonkers [0] (
22:12:12 Join ashridah [0] (
22:12:41 Join Renko [0] (
22:13:20Bonkersso I'm trying to use the firmware flasher for the first time, and to live dangerous, I'm using the bleeding edge build from an hour or so ago, I flashed the firmware on my recorder, it boots fine with the default firmware, now I want to load in the rockbox firmware, so I had to create the .ucl file with "uclpack -nrv2e -9", after I load that file into the firmware and try to boot it, I just get a blank screen, although I can still hold F
22:13:20Bonkers1 to boot, what might I have done wrong?
22:13:42BonkersI created the .ucl file right from the ajbrec on the recorder, so it's the same version
22:16:01Camilopreglow, has anyone got the mac instructions used yet?
22:16:21ACcamilo: libFalc
22:17:20ACsee soldifre.c in libFLAC dir
22:17:32BonkersI used the uclpack that came with libucl, is that wrong?
22:17:37ACgives about 3-4 % speedup
22:18:22CamiloAC, thanks, I was struggling with the mad lib
22:18:32Camilois it really as low as 4%?
22:18:49pabsHCl: will using hte remote allow us to get around the one-button limit?
22:19:11pabsHCl: ie can you have a button on the remote and a button on the unit at the saem time?
22:19:33pabsotherwise its impossible to finish the first level in bionic commando
22:20:12ACCamilo: yeah.. atm only 3-4 %speedup, but preglow is working on it
22:21:54Camilohow much are 2nd hand 120s going for?
22:21:58CamiloI have a H340
22:22:19Bonkersnobody knows about creating ucl files?
22:22:33HClpabs: dunno.
22:23:50preglowCamilo: i'm using them all the time
22:24:44Camilodoes the cpu have an instruction cache? That unrolled code in the libflac looks large
22:24:49 Quit Sucka (Read error: 60 (Operation timed out))
22:25:09HClyes, it does.
22:25:22Camilois it shared with data or just for instrs?
22:26:28preglowCamilo: just instr cache
22:26:40 Join mecraw [0] (~mecraw@
22:26:49preglowCamilo: and it's not particularily huge, a mac instruction is two words
22:27:32 Join stevenm [0] (
22:28:03Camiloit's cutting edge stuff, really interesting
22:28:08stevenmHello people
22:28:33Camilodo you know what the original firmware uses? Presumably the mac instructions stand out in a disassembly ;-)
22:29:43 Join [IDC]Dragon [0] (
22:30:09preglowCamilo: i have no idea, but you can be certain they are used
22:30:31preglowcoldfire.c is about to be replaced by the new and improved coldfire.S anyway!
22:30:32preglowyay me
22:30:58stevenmpreglow, did you say there was a way to remove sampling noise using interpolation and possibly a .. cubic spline?
22:31:08preglowstevenm: linear should be enough for starters
22:31:11Camilogot to go,.. but I look forward to seeing that stuff in CVS :)
22:31:20preglowCamilo: it'll be in cvs tomorrow, possibly
22:31:36CamiloAC thanks for the flac pointer
22:31:39Camilocheers preglow
22:31:47 Quit Camilo ("Chatzilla 0.9.67 [Mozilla rv:1.8a6/20050111]")
22:31:47ACCamlio: np
22:35:29 Join [IDC]Dragon2 [0] (
22:41:53preglowAC: i can't see anything obvious, i'm afraid
22:42:04preglowAC: it's very dissimilar to everything else i've seen
22:42:22pabsHCl: zelda works great
22:43:03ACpreglow: hmmm.. i will send the author of libwavpack a mail.. maybe he can help...
22:44:45preglowAC: do so, and mention to him the mac facilities we have, he might know of a place or two where you can benefit from it
22:45:05pabsHCl: except i just accidentally shut down when i was trying to get my sword :(
22:45:27preglowHCl: you got a fixed version of rockboy lying around?
22:45:43stevenmpreglow, how do you actually do a linear interpolation in this case?
22:45:48HClpreglow: not at the moment.
22:46:01preglowHCl: the one i have hangs on exit and stuff like that
22:46:26stevenmpreglow, I have a 'ghetto' lowpass filter to get rid of the noise for now, but I guess not making it in the first place would be ideal. Question is, which is less CPU usage?
22:47:10preglowstevenm: well, okay, what format is your samples in?
22:47:12preglowstevenm: 16 bit?
22:47:16stevenmpreglow, yes
22:47:39 Join Sucka [0] (
22:48:13preglowwell, ok, interpolated_value = sample[x] + (sample[x + 1] - sample[x])*fractional
22:48:27preglowwhere fractional is the position you are at between the samples
22:48:42preglowahh, and you need a shift after the mul
22:49:16 Join lImbus [0] (
22:49:19lImbushi all
22:49:36lImbuswahoo. 57 users.
22:49:55HClpreglow: the rockboy.rock on my ftp works at the moment..
22:50:31HClnew button driver works great too - it can detect movement and the A/play button being pressed at the same time
22:50:39HClmakes mario much more playable
22:50:45preglowHCl: excellent!!!!
22:52:06HCli became super mario and i can shoot balls now
22:52:51 Quit [IDC]Dragon (Read error: 110 (Connection timed out))
22:53:09HClturtle shells explode.
22:53:12HClokay, i didn't know that
22:53:28ashridahyeah, if you stomp on them
22:53:31HClnew button driver helps such a great deal
22:53:34 Nick [IDC]Dragon2 is now known as [IDC]Dragon (
22:53:41HClmakes mario fully playable aside from shooting
22:54:06ashridahfully? as in you have diagonals?
22:54:22HClwell, you don't need diagonals for mario
22:54:24HClonly with jumping
22:54:30ashridahyes you do
22:54:30HCland you can do jumping independantly from movement
22:54:38ashridahever tried a crouchslide? :)
22:54:45HClno o.o
22:55:10*[IDC]Dragon drowns in old Archos'es
22:56:29stevenmWhat about the remote buttons for rockboy ?
22:56:36ashridahHCl: there's always the odd spot only a small mario can get to, but some of them can be entered by a super mario, by taking a run up, jumping, and as you hit the ground, crouching and sliding forward
22:56:49HClyou only need down for that...
22:56:52HCli think.
22:57:46ashridahperhaps, but the temptation to use a diagonal is there
22:58:14preglowyou only need down for that
22:58:20preglowthat's how i used to do it
22:58:51HClmario level 1 cleared
22:59:06ashridahwhat's the framerate these days?
22:59:17HClnear full speed
22:59:34 Join geoff_o [0] (
23:00:05ashridahhow'd you end up getting around the slowness of the lcd?
23:00:19lImbus[IDC]Dragon, listening ?
23:00:22 Quit Stryke` ("Friends don't let friends listen to Anti-Flag")
23:00:35lImbusI's about one more old archos, my old shielded ondio
23:00:41 Quit jyp ("poof!")
23:00:43lImbusI got it back today from my co-worker
23:00:48[IDC]Dragongot no Ondios today
23:01:00Bonkerswhat device are you running mario on?
23:01:18HClashridah: its not very slow o.o
23:01:26lImbus[IDC]Dragon, it's about the one I am supposed to send you in the future
23:01:38ashridahHCl: ah, so it was just the method you started out with :)
23:01:50[IDC]Dragonyes, just making a little fun
23:01:51ashridahBonkers: iriver H1xx
23:01:54HClmy very first version used drawpixel and clearpixel
23:01:58lImbusthe problem my coworker had is, that it seems impossible to get smd-soldering iron (or even tips) in aachen
23:02:02 Join Bagder [0] (
23:02:13Bonkerswhat do you need SMD tips for?
23:02:25lImbusto desolder the radio board
23:02:35lImbusand NOT damage it whilst
23:02:36Bonkerswhy not just use some solderwick?
23:02:50RenkolImbus, hot air?
23:03:04Bonkersor hot air works too and can be done with like a hair dryer and some other mods
23:03:16***Saving seen data "./dancer.seen"
23:03:23lImbusI got a gaz-soldering iron with a hot-air nozzle :-)
23:03:23*Renko strokes his weller hot air station
23:03:26amiconnI'm puzzled... Why does Linus' Makefile adjustment for .S files magically fix the ton of inline warnings from libFLAC?
23:03:35Renkowhich isnt actually his :(
23:04:36BonkersI solder all my SMD stuff with a $6 iron from rat shack
23:05:51Bonkersdown to QFP32 I've done
23:06:28Renko56pin ssop?
23:06:38Bonkerswhat's ssop look like?
23:06:53Bonkersis that the one that's just smaller than QFP? I have one, I haven't attempted it yet
23:07:15Renkoit's a dil
23:07:24Renkoo.5mm pitch tho
23:07:54Renkocan be done with a $6 iron but I dont fancy it :)
23:08:13Bonkersheh, I have no problem with it ;)
23:08:52Renkoi'll have to not have a problem once i leave uni this year, heh
23:09:39*HCl slaps his university internet connection with a rubber platypus
23:10:22*preglow hugs his
23:10:39HClwell, its fine, but we just had a kiddie on our net
23:10:58HClwho claimed the ips of all our core systems, causing all computers to lose internet
23:11:01Bonkerslooks like LQFP32 is 0.8mm pitch, so SSOP would certainly be getting tough
23:11:11HClhis internet connection was terminated immediately though, so we should have wireless back in a minute
23:12:46Renko*it's tsop that's 5mm pitch
23:13:39Bonkers5mm? I can do that with a lighter ;)
23:13:58Renkoeh 0.5 8-)
23:13:59 Join murphy| [0] (
23:23:02 Quit murphy| ()
23:23:45Bonkersanyone know about the id3db on the recorder? I'm using CVS from a few minutes ago, I generated the db, I got .rockbox/rockbox.id3db on my recorder now, yet when I browse by ID3, I get nothing
23:24:49 Quit frank_ ("Leaving")
23:26:32*stevenm bashes his head against a wall as he tried to find the cause of a Floating Point Exception in integer-only code.
23:27:17lImbuspreferably on an integer only processor
23:27:30lImbusBonkers, I got the same prob and tried a lot.
23:27:44lImbusI succeeded once, but don't know how.
23:27:49lImbusI should retry this tho
23:27:58BonkerslImbus: you do anything special?
23:28:12amiconnBagder, [IDC]Dragon: Anything that speaks against committing rockboy?
23:28:32BagderI'm not up to speed with recent events, so I cannot tell
23:28:39BagderI just got back home
23:28:47lImbusBagder, wlcome back :-)
23:28:52amiconnWelcome back :)
23:28:54*[IDC]Dragon spots Bagder
23:29:02Bagderwas fine
23:29:06Bagder20C+ and sun
23:29:10amiconnBagder: The remaining simulator issues are fixed meanwhile
23:29:15Bagderhere we have -14C and snooooow
23:29:17preglowpfaw! i prefer -15c and snow
23:29:36Bagderamiconn: nice!
23:29:44[IDC]Dragonamiconn, I didn't follow Rockboy, can't tell
23:30:07amiconnBagder: There's one thing remaining with the old-gcc Win32 crosscompile, it generates 2 warnings about implicitly declared functions
23:30:16amiconnThat doesn't happen here
23:30:22Bagderamiconn: I'm generally *for* committing though
23:30:27lImbusBonkers, think on almost all my tries, I switched on the unit, switched to "id3-browse", then did change back to "supported" as soon as I saw it did not work. I THINK the one time where it worked, I switched on the unit while it already was in id3-mode
23:30:33[IDC]DragonBagder: the OndioFM daily is somehow broken, I heared
23:30:44lImbusbut not reproduceable neither
23:30:49Bagder[IDC]Dragon: I think it is due to the flash full error
23:30:59Bagderbut I haven't investigated much
23:31:08[IDC]Dragonperhaps, I also fail to do "make zip"
23:31:15*Bagder catches up on a few thousand emails
23:31:17amiconnHCl: I have a brand-new little icon for .gb / .cgb Dunno if you like it...
23:32:02[IDC]Dragonamiconn, are you familiar with the powermgmt code?
23:32:24amiconnBagder, [IDC]Dragon: For rockboy on archos, I need an 'overlay' technique to load it into the mp3 buffer.
23:32:29[IDC]DragonJerrys patch disturbs the Ondio
23:33:11amiconnI just chose the extension .ovl. Maybe you have a better suggestion?
23:33:13[IDC]Dragonamiconn, is that prepared for 8MB as well?
23:33:19amiconn[IDC]Dragon: yes.
23:33:39Bagder.ovl works for me
23:33:42amiconnThere is an that uses the memory size definition
23:34:19amiconnI also need to know how to make this new file included by
23:34:19 Quit AC ("CGI:IRC (EOF)")
23:35:06BonkerslImbus: now that you mention it, I remember that it has to be on when you turn it on the first time so it can allocate some memory somewhere
23:35:11amiconnFor iriver and the simulators, there's just rockboy.rock, but the archos recorders need both rockboy.rock and rockboy.ovl in .rockbox/viewers/
23:35:34BonkerslImbus: and now it works here
23:35:40HClamiconn: i'm probably severely breaking things with the grayscale support at the moment..
23:35:41Bagderamiconn: I can fix that once you've commited
23:36:01HCloh, nice, icon, :D any icon is better than none :p
23:36:08amiconnHCl: Of course. Because of that I cannot commit your latest
23:36:18HCli know
23:36:24HCli was going to have it in a nice define
23:36:32HClbut part of it refused to compile with the defines
23:36:55amiconnI'll commit the plain old non-dynarec first, then work in your dynarec efforts and commit again
23:37:23 Quit Sucka (Read error: 110 (Connection timed out))
23:37:26HClat the moment, i'm slightly starting to worry to whether dynarec will help
23:37:27amiconnThe greyscale support has to wait until greyscale for iriver is in cvs too.
23:37:47HCli'm thinking that an assembly interpreter core might be fast enough
23:37:53HClsince its already at near full speed
23:37:53amiconnThen greyscale support must be added to the simulators as well
23:38:06HCldespite my changes to the interpreter to make it compatible with dynarec
23:38:12HClwhich are actually making it somewhat slower
23:38:16HClthough i can't say to what extend
23:41:33HCli swear
23:41:38HCl ; 's behind an IF statement
23:41:39amiconnWell, I could make good use of some speedup
23:41:41HClshould produce a warning.
23:41:47HClthat an if statement isn't doing anything
23:41:57HClamiconn: yea, you're right
23:42:01HCldidn't consider archos
23:42:06lImbusBonkers, nice info, thanks.
23:42:17HCloh, i rewrote the button driver today
23:42:21HClit works much nicer now
23:43:08[IDC]Dragonamiconn: how did you solve the 2-stage viewer dilemma? (playing .gb files)
23:43:27amiconnThat's easy.
23:43:49amiconnrockboy.rock is just the loader, which loads rockboy.ovl and then passes all parameters
23:44:07Bonkersdoes rockboy do anything on the archos recorder?
23:44:26amiconnyes, it is working on then as well
23:44:43[IDC]Dragonamiconn,I was rather thinking about a generic .ovl viewer
23:44:47amiconn(but slowly, and the low-res lcd makes it somewhat hard to identify things)
23:45:03BonkersI saw some passing conversation before, you're doing linear interpolation?
23:45:07[IDC]Dragon(but you could check for the extension)
23:45:19amiconnBonkers: No, just dropping pixels
23:45:26Bonkersnearest neighbor?
23:46:23amiconnIt simply drops every 2nd line vertically; the horizontal scaling does 160->112 by a bresenham-like algo
23:46:34Bonkersand this is in CVS now?
23:46:42amiconnI'm about to commit
23:46:51Bonkerscan't wait to try it out
23:47:12preglowdo i need to put anything special in the source file headers to have the cvs version thing inserted?
23:47:48Bagderpreglow: just copy an existing header
23:48:03Bagder$Id:$ gets replaced by cvs
23:49:38preglowdoes anyone happen to have flac files they know utilizes lpc blocks of all orders? :V
23:50:23preglowBagder: <- like this, yes?
23:50:50Bagderpreglow: yeps
23:54:36 Quit lolo-laptop ("Client exiting")
23:54:43HCli don't seem to be able to check the hold button :(
23:54:59HCldoes anybody know how? i'm trying to use button_hold() but it seems to return false always
23:55:56HClamiconn: hold on, let me give you my new button driver
23:55:58HClbefore you commit
23:56:39HClgrab the new button driver function out of that, and remove the reference to lcd_mode
23:56:55HClit makes mario so much more playable

Previous day | Next day