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 2009-01-19

00:02:07 Join Comokanu [0] (n=5c0ce2b3@gateway/web/cgi-irc/
00:02:13 Quit Comokanu (Client Quit)
00:05:49 Quit Horscht ("Verlassend")
00:06:28 Join Horscht [0] (n=Horscht@xbmc/user/horscht)
00:06:41 Join solexx [0] (
00:10:32gartraland i cant get a sim build too work
00:11:31Bagdergartral: what happens?
00:12:51gartralthe sim wont even load
00:13:10Bagdercan you be more specific?
00:13:24Bagderthat dev environment do you use?
00:14:03gartrali built a sim build, and i ziped the simdisk dir, and transfered to my comp and unzipped into the local simdisk, replacing what was there, and now the sim is busted
00:14:26saratogatry make install instead of make zip
00:14:35saratogato rule out other problems
00:14:36gartrali use the build enviroment that JdGordon has, as its his system i build from
00:14:53 Quit domonoky (Read error: 54 (Connection reset by peer))
00:14:56 Join Rob2223 [0] (
00:14:59saratogayou'll probably need to run it on his system then?
00:15:06Bagdergartral: well, without more specific details it sounds as if your two systems are similar enough then
00:15:22Bagderare not
00:15:27Bagderdarnit I should go to bed instead
00:15:30gartralwhat do you mean?
00:15:53Bagdersomething prevents it from running on your system
00:16:09Bagderpossibly because your system differs from jdgordon's somehow
00:16:19gartrali figured that
00:16:48gartralbut why would it be different if all im changing is the .rockbox directory?
00:16:59Bagderyou tell us!
00:17:17BagderI'd use strace or gdb to find out on a linux box
00:17:21gartrali dont understand what you mean
00:17:52saratogatheres probably some difference between your systems like OS versions, bitness, etc
00:18:05gartralwell, my system is windows, his is linux, i know i cant make the executable work, but why would the base files fail?
00:18:23saratogabacause they're compiled for linux and you're not running linux
00:18:43 Quit BHSPitLappy (Remote closed the connection)
00:19:12saratogaif you want to run the sim on windows you'll need to compile it for windows, unless Jdgordon already did that for you
00:19:40amiconnAnd you'll obviously need sdl.dll as well
00:19:44gartralwell, as it doesnt work, ide say he didnt
00:19:52 Quit solexx_ (Read error: 110 (Connection timed out))
00:19:55gartralalready there
00:20:47 Quit n17ikh (Read error: 60 (Operation timed out))
00:21:12gartralhow do i cross compile for windows
00:22:28saratogai think the easiest way is to setup the dev environment on windows and run it in there
00:23:01gartrali just dont understand why the applications and such need compiling for windows if there for rockbox...
00:23:06linuxstbgartral: Or you could just download a binary for Windows - the same UiSimulator wiki page has links.
00:23:27Bagderor just build and run it on linux...
00:23:32linuxstbgevaerts: That wiki page also has a useful description of what the sim is...
00:23:42linuxstb(sorry, I meant gartral)
00:23:45gevaertslinuxstb: thanks :)
00:23:52saratogawindows can't run linux programs
00:24:01saratogathis is the answer to your question
00:24:03linuxstbgevaerts: You're welcome. Now please read it.
00:24:16gevaertslinuxstb: I know how to build windows sims on linux!
00:25:41 Join casainho [0] (
00:26:01casainhohello :-)
00:26:34casainhodoes anyone knows if there is same target player that uses SD card with SPI? - I am looking for SPI drivers of SD cards...
00:29:01gartralok.. ill just grab a simbuild from the site, as i really dont feel like breaking my build enviroment a third time
00:29:32 Quit timc`` (Connection timed out)
00:31:11 Join timc`` [0] (n=aoeu@
00:31:22 Join BHSPitLappy [0] (n=BHSPitLa@unaffiliated/bhspitmonkey)
00:31:48 Quit Rob2222 (Read error: 110 (Connection timed out))
00:31:54 Join tajlero [0] (
00:35:12 Quit dfkt ("-= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.")
00:37:41 Quit casainho ("ChatZilla 0.9.84 [Firefox 3.0.5/2008121622]")
00:38:11 Join DerPapst [0] (
00:38:19DerPapsthey all :-)
00:39:51tmztkugel: what svn revision is preferred for building mkamsboot to flash fuze and retain the ability to boot OF?
00:40:12kugeltmzt: most recent?
00:40:34tmztthere are no issues with key detection that would likely brick the fuze?
00:41:25 Quit ibseco (Read error: 145 (Connection timed out))
00:41:48gartralok, this worrisome, the files seem too play fine in the sim
00:42:21DerPapstJdGordon|afk (since you did that or any other one who cares): dunno if that is correct but in lamp.c the default option in the switch statement for targets with a color display the !SYSEVENT feels wrong... shouldn't it be 'if (IS_SYSEVENT(button))'?
00:43:26DerPapstanyway, free time's already up again. see you all some other time ;-)
00:43:29 Quit DerPapst ("iPodLinux rocks!")
00:44:31 Quit Slack_ (Read error: 60 (Operation timed out))
00:45:57 Join HellDragon [0] (
00:48:36gartralok, it had too play 3 songs, but it did freeze and quit
00:48:55gartralbut without stating an abort address
00:51:24 Quit ender` (" If you eat a live toad first thing in the morning, nothing worse will happen all day long. -- California saying; To you or")
00:52:35 Quit jhMikeS (Read error: 110 (Connection timed out))
00:54:42 Join Slack_ [0] (
00:59:58 Join jhMikeS [50] (n=jethead7@rockbox/developer/jhMikeS)
01:02:24gartralthe dap aborts as soon as the userfl buffer fills
01:02:51 Quit MethoS- (Remote closed the connection)
01:03:05 Join LambdaCalculus37 [0] (n=rmenes@rockbox/staff/LambdaCalculus37)
01:05:19 Quit blithe (Read error: 60 (Operation timed out))
01:06:29 Join blithe [0] (
01:06:50gartraland all the buffers fill almost instantly on just that album it sems, in bothsim and real
01:08:26 Quit itcheg (" ajax IRC Client")
01:09:32 Join itcheg [0] (i=62db4767@gateway/web/ajax/
01:14:28 Join japc_ [0] (
01:15:12 Quit timc`` (Read error: 104 (Connection reset by peer))
01:15:24 Join timc`` [0] (n=aoeu@
01:17:29 Quit itcheg (" ajax IRC Client")
01:17:40 Quit japc__ (Read error: 145 (Connection timed out))
01:18:16 Join Willwolfe [0] (
01:19:05 Join itcheg [0] (i=62db4767@gateway/web/ajax/
01:21:57 Quit flydutch ("/* empty */")
01:22:47gartralnot often my hunches work, but i just refreshed the database and now, playing the play list that caused the abort suddenly doesn't...
01:23:19 Quit nuonguy ("This computer has gone to sleep")
01:24:53Unhelpfulkugel: how'd i break test_boost?
01:26:13UnhelpfuljhMikeS: not frustrations, so much as wondering which wrappers we can remove. i guess the cache function ones seem to be more about allowing direct calls to inline versions on some targets, and using the API on others, than about hiding the API in general?
01:27:56Unhelpfulamiconn: is there a separate 32x32->32 multiply on arm? if that's just umull + discard bottom half, i'd like to keep the present code on arm, since it has 7 bits more precision and no actual shifts
01:28:19amiconnOf course. Just 'mul'
01:28:24 Join gewuerzwiesel [0] (n=gewuerzw@unaffiliated/gewuerzwiesel)
01:28:26 Join nibbler [0] (
01:28:41amiconnSigned and unsigned are the same if you do n*n->n
01:28:59amiconn'mul' is 1 cycle faster than smull/umull
01:30:11Unhelpfulok... maybe i need to see what the practical limits are if we move everything over 8 bits
01:30:33 Quit itcheg (" ajax IRC Client")
01:30:34 Quit Slack_ (SendQ exceeded)
01:30:51 Join itcheg [0] (i=62db4767@gateway/web/ajax/
01:32:03Unhelpfulif we use mul, there would be a shift though, right? the shift comes for free with umull, since the output is in two registers.
01:32:16 Join Slack_ [0] (
01:33:17amiconnThe shift might also be for free, depending on how the mul result is processed further
01:33:42Unhelpfulit would be >>24 after the mul
01:34:18amiconnYeah, it depends on the instruction that processes the 'mul' result
01:34:32saratogamany arm instructions can shift their inputs, so the shift is probably free
01:35:01amiconnThey can shift one of their inputs
01:35:23saratogaUnhelpful: any interest in a jpeg decoder once this is all done?
01:35:28 Join Rob2222 [0] (
01:35:48Unhelpfulwell, the next thing would be putting it into LCD_RGBPACK_LCD, in the PF case, or into dithering, in core...
01:36:24UnhelpfulLCD_RGPACK_LCD is going to be shifting it, and i'm guessing shift can't shift before it shifts?
01:37:14 Quit gromit` ("Coyote finally caught me")
01:37:33 Join gromit` [0] (
01:38:36 Part gewuerzwiesel
01:39:57 Quit tmzt (Read error: 110 (Connection timed out))
01:42:24 Quit tessarakt ("Client exiting")
01:42:37 Quit itcheg (" ajax IRC Client")
01:42:57 Quit PaulJam_ (".")
01:45:49kugelUnhelpful: well, you didn't apply your plugin api rework to it
01:46:26gartraljpeg? that'd be nice >.>
01:47:22 Join Horschti [0] (n=Horscht@xbmc/user/horscht)
01:48:23 Quit nibbler (Read error: 110 (Connection timed out))
01:49:21 Part gartral
01:49:57***Saving seen data "./dancer.seen"
01:50:35 Join AndyI [0] (i=AndyI@
01:51:06 Join itcheg [0] (i=62db4767@gateway/web/ajax/
01:51:56 Quit gromit` (Read error: 60 (Operation timed out))
01:52:20 Quit Rob2223 (Read error: 110 (Connection timed out))
01:52:38kugelUnhelpful: the fix is easy
01:58:41 Quit itcheg (" ajax IRC Client")
01:59:14 Join gromit` [0] (
01:59:46 Quit LambdaCalculus37 ("Ka-chunka")
01:59:49Unhelpfulit should be, yes. must've been merged after my first patch?
02:02:06 Join itcheg [0] (i=62db4767@gateway/web/ajax/
02:02:24Unhelpfulamiconn: the dither math is like this: r = (31 * r + (r >> 3) + delta) >> 8;
02:02:33 Quit itcheg (Client Quit)
02:03:30Unhelpfulso, if it can calculate the 31*(r>>24) without destroying r, the r>>3 becomes r>>27... i'll probably need to either time it or go over the asm to see if this is faster than umull?
02:05:05 Quit tajlero ("Leaving")
02:05:06 Quit Horscht (Read error: 110 (Connection timed out))
02:07:11 Join nuonguy [0] (
02:12:21Unhelpfuli'll set up a few defines to make the shift size switchable, so it won't be hard to test both versions on ARM
02:14:01 Join axionix [0] (
02:14:01kugelUnhelpful: right
02:14:29 Join Soap_ [0] (
02:15:12 Quit BHSPitLappy (Read error: 60 (Operation timed out))
02:15:34 Quit culture (Read error: 110 (Connection timed out))
02:16:08 Quit nuonguy (Read error: 148 (No route to host))
02:18:58 Join Rob2223 [0] (
02:19:28 Join gromit`` [0] (
02:19:36 Quit Thundercloud (Remote closed the connection)
02:19:54Unhelpfulkugel: that should fix it. i wonder if there are any docs i should fix?
02:20:13 Join XavierGr [0] (n=xavier@rockbox/staff/XavierGr)
02:21:51kugelUnhelpful: not that I know, but you might want to put test_boost into a subdir (right now it's in rocks/ not in rocks/apps) while you're at fixing it
02:23:11saratogamicroboosting does give a nice battery improvement on PP for MP3 according to the FS entry
02:23:12Unhelpfuli already committed the global plugin API fix... those basically just involve removing an argument from plugin_start and removing any declaration of a plugin_api. i ported almost everything with sed. ;)
02:24:15Unhelpfulperhaps just define a SCALER_SHIFT and key the rest of them from that?
02:24:16saratogai bet Coldfire gets an even bigger increase
02:24:34 Join cool_walking_ [0] (i=cb3b81c3@gateway/web/ajax/
02:27:43 Quit axionix_ (Read error: 110 (Connection timed out))
02:28:43 Join tmzt [0] (n=tmzt@
02:29:21 Quit gromit` (Read error: 110 (Connection timed out))
02:34:58 Quit Willwolfe ("ChatZilla 0.9.84 [Firefox 3.0.1/2008070208]")
02:36:11 Quit Rob2222 (Read error: 110 (Connection timed out))
02:39:46jhMikeSUnhelpful: It would otherwise need nasty #ifdef's in every instance as rb->cache_fn() or just cache_fn(). Really, I'd prefer rb->cache_fn() always and save the nonsense.
02:40:29 Join tajlero [0] (
02:40:36UnhelpfuljhMikeS: maybe... but understanding the reason, and knowing it's not a mere convenience wrapper, makes me not really have a complaint against it any more
02:41:22 Join tseymour [0] (
02:41:36 Join BHSPitLappy [0] (n=BHSPitLa@unaffiliated/bhspitmonkey)
02:42:44 Quit cool_walking_ (" ajax IRC Client")
02:44:39jhMikeSUnhelpful: I think it's an unneeded complication in the system for no real benefit. I'm guilty of creating it but it just seems rather pointless...sort of a pathological avoidance of changing too much, which I think I've cured myself of nowadays.
02:46:02Unhelpfuli guess it depends on how large the function call overhead is, and how often they're called?
02:48:39jhMikeSthey aren't called often and full operations like that can be rather slow anyway
02:52:37 Quit Seed ("cu, Andre")
02:53:05 Join Seed [0] (
02:53:27 Nick fxb is now known as fxb__ (
02:58:21TheSkunkManI'm trying to use the SimpleFirmwareCrack, but how do I set the input file? Shouldn't it be like: g++ decode_light.cpp -i file ?
02:59:45 Join cool_walking_ [0] (i=cb3b81c3@gateway/web/ajax/
03:00:31 Join Rob2222 [0] (
03:00:56kugelsaratoga: wow, 23.5h on mp3
03:01:13kugelslowly, rockbox runtime on sansas get impressive :)
03:02:08kugelbut I still think 20MHz is too low for the UI (especially text editor), even though it's not as worse as I thought
03:04:05 Quit kugel ("ChatZilla 0.9.84-rdmsoft [XULRunner]")
03:10:03 Quit tajlero ("Leaving")
03:10:18 Join tajlero [0] (
03:10:39 Quit Slack_ (Connection timed out)
03:11:40Unhelpfulamiconn: looks like i'll need to change the rounding scheme for >> 24 - rounding the reciprocal up, rather than to nearest, is causing overflows even at sizes like 500x500
03:14:42 Quit tyfoo ("Carpe diem")
03:16:30 Quit axionix (Read error: 60 (Operation timed out))
03:18:05 Join JimCokos [0] (n=442e3aa5@gateway/web/cgi-irc/
03:18:20 Quit Rob2223 (Read error: 110 (Connection timed out))
03:18:50 Join RoC_MasterMind [0] (
03:20:38 Join axionix [0] (
03:21:04 Quit JimCokos (Client Quit)
03:22:23 Join Slack_ [0] (
03:23:30Unhelpfulok... it can work at reasonable sizes, at least. starts to fail at substantially less unreasonable sizes than the >>32 version does.
03:32:14 Join Darksair [0] (n=user@
03:36:38 Quit Slack_ ("Ex-Chat")
03:42:34 Join Rob2223 [0] (
03:44:25Unhelpfulamiconn: which is going to be worse on sh-1 and coldfire... a 32-bit divide, or a 32x32->64 multiply (with the undertstanding that that may be 64x64->64 in reality)
03:44:58 Quit m0f0x ()
03:49:52 Quit RoC_MasterMind ("Leaving")
03:49:58***Saving seen data "./dancer.seen"
04:00:53 Quit Rob2222 (Read error: 110 (Connection timed out))
04:07:10 Quit timc`` (Read error: 145 (Connection timed out))
04:07:54 Quit tajlero (Read error: 60 (Operation timed out))
04:17:37 Join Barahir_ [0] (
04:20:49saratogaUnhelpful: its hard to imagine theres a target where divides are faster then multiplies of any size
04:21:37Unhelpfulsaratoga: even multiplies two or four times as wide as supported by the widest multiply instruction?
04:22:03 Quit Darksair ("People who are zhuangbility want to show their niubility but only reflect their shability.")
04:22:40saratogaUnhelpful: I would expect so
04:22:48saratogabuilding up larger multiplies is pretty quick
04:23:02saratogawhereas even hardware division is typically quite slow
04:25:32Unhelpfuli know divisions are going to have to be done iteratively... but it takes a fair number of multiplies to get a 32x32->64 or 64x64->64 when you only have 16x16->32, which is what i understand sh has
04:25:47 Join xSlack [0] (
04:26:50 Join tajlero [0] (
04:28:02saratogaARM only has an 8 bit multiplier
04:28:11saratogawell armv4
04:28:37saratogaand it manages to do 32x32->64 multiplies quite quickly [6 clocks]
04:28:48Unhelpfulgood gravy... it may be wise for me to use the shift24 version on some ARM as well :)
04:28:48 Join timc`` [0] (n=aoeu@
04:29:48saratogayes armv4 is really shitty
04:30:15 Quit Barahir (Read error: 110 (Connection timed out))
04:30:43*Unhelpful didn't know about the ability of arm to apply shifts to arguments of other operations, either
04:31:08Unhelpfulso i'll be testing if the >>24 version without the 32x32->64 multiply is faster on arm as well
04:31:37saratogaif it saves a multiply i would expect it to be
04:32:21saratogathats pretty much what all codec optimization on arm is, trying to figure out how to not multiply numbers
04:33:01Unhelpfulthere's still a multiply. amiconn said 32x32->64 costs a cycle more than 32x32->32, so the real question is whether we end up losing that cycle with the need to shift the value in the shift24 version
04:33:18saratogai wouldn't worry too much about saving one clock
04:33:25saratogaunless you're going to do that multiply a whole lot
04:33:48Unhelpfulper output pixel in the scaler.
04:34:00saratogathats nothing
04:34:09saratogayou'll never have even a million output pixels
04:34:46Unhelpfulheh, true
04:34:56saratogaat that point you're probably better off unrolling the loop in ARM asm to save all clocks wasted on the branch and lw instructions
04:35:52 Quit AndyI (Read error: 145 (Connection timed out))
04:36:09Unhelpfulwell, 80MHz PP scales a 2560x1600 image down for sliding_puzzle in 16s
04:36:37 Part tseymour
04:37:14Unhelpfulwould unrolling the loop in asm really be much more worthwhile than unrolling it in C?
04:37:25saratogadepends how much faith you put in gcc
04:37:46saratogai've beaten it by a lot, and also found times where i could do no better
04:38:00 Quit jhulst (Connection timed out)
04:38:07Unhelpful...i've had it produce broken code when hand-unrolling a loop in C :/
04:38:08saratogafor very tight loops its often worth trying since the effort is small and saving even a few clocks on unrolling can make a big difference
04:39:01saratogabasically look for anytime you do a lot of "load, mult, store" loops, as those can often beat GCC by unrolling 2 or 4x, then replacing the lw with load multiple ops
04:39:43saratogaparticularly for PP, since load multiples are almost twice as fast as load singles
04:39:52 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
04:40:26*Unhelpful would think that gcc would know how to do that itself...
04:42:15Unhelpfuli guess everything ARM has decent memory, as well, ie, nobody will kill me if a loop unroll adds a few bytes...
04:47:35saratogaunrolling is tough for compilers, since it wastes a lot of space and its not always clear if its worth it at compile time
04:47:44saratogaand gcc for arm is not the smartest compiler
04:48:12Unhelpfuli'm guessing that gcc is not able to turn that into a shift+<otherop>, and is adding an explicit shift, because the shift24 version is slower than the shift32 version with its free shift
04:48:34Unhelpfulon e200 - PP 80MHz, right?
04:49:12 Quit fdinel ("Miranda IM! Smaller, Faster, Easier.")
04:49:34 Join AndyI [0] (n=pasha_in@
04:49:37 Join blkhawk- [0] (
04:50:04saratogaUnhelpful: yes
04:50:11Unhelpfulwell, in any case, i seem to have a working scaler that never needs 32x32->64 mul, and i'd like to use that version wherever we don't have that operation in hardware... sounds like everything besides arm>4?
04:50:32saratogaUnhelpful: its pretty easy to objdump your own code and see what gcc is doing
04:51:15saratogajust add whatever that compiler switch that leaves in the debug info to your makefile, and then arm-elf-objdump it
04:51:17Unhelpfulyes, but i don't know much at all about arm asm... so it was also pretty easy to count ticks, then make my e200 load a huge image while nothing else is happening
04:51:43saratogai don't really know anything about it either, but if you're just mutliplying, adding, shifting, loading and storing you ddon't need to know much
04:58:16 Quit miepchen^schlaf (Read error: 110 (Connection timed out))
04:59:41 Join intrados [0] (
05:03:51Unhelpfulok... so for the moment, i want to use this code #if defined(CPU_ARM)... are there any specific CPUs we use that are ARM4 that i want to use the less precise version on?
05:05:47 Quit blkhawk (Read error: 110 (Connection timed out))
05:06:34 Nick blkhawk- is now known as blkhawk (
05:09:18 Quit XavierGr ()
05:26:02Unhelpfulamiconn: here's what i have so far. the quality is good with SC_SHIFT == 24 up to reasonable sizes, but rounding/division errors start to happen a bit earlier. SC_SHIFT == 32 times faster on my e200, so i'm using that if CPU_ARM is defined.
05:28:40 Quit Acksaw (Read error: 104 (Connection reset by peer))
05:29:02 Join Acksaw [0] (
05:36:40 Quit saratoga ("CGI:IRC (Ping timeout)")
05:49:27 Join saratoga [0] (n=9803c6dd@gateway/web/cgi-irc/
05:49:53 Join grndslm [0] (
05:50:00***Saving seen data "./dancer.seen"
06:03:40 Quit Tuplanolla (Read error: 145 (Connection timed out))
06:06:50 Join tim__ [0] (n=aoeu@
06:10:39 Quit timc`` (Read error: 54 (Connection reset by peer))
06:15:54 Join n17ikh|Lappy [0] (n=n17ikh@
06:16:49 Nick n17ikh|Lappy is now known as n17ikh (n=n17ikh@
06:17:13 Nick n17ikh is now known as n17ikh| (n=n17ikh@
06:17:18 Nick n17ikh| is now known as n17ikh (n=n17ikh@
06:20:48 Join os2user [0] (n=chatzill@
06:22:44 Quit kadoban (Remote closed the connection)
06:22:59 Quit jhulst (Read error: 54 (Connection reset by peer))
06:30:15 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
06:41:10 Quit Prot (Read error: 113 (No route to host))
06:44:11 Quit AndyI ()
06:48:50 Join unstable [0] (
06:49:09 Join __lifeless [0] (n=lifeless@
06:49:13unstablehmm, the Sansa e280 of sucks. Anyone know how I can just see file names?
06:51:12Unhelpfulunstable: files in the main menu. you'll have to set it to show all files if you want to see the "music" directory, since the OF marks that hidden on startup
06:52:41unstableUnhelpful: "files"? in the main menu? I don't see a files option.
06:53:01Unhelpfulit's the first item in the rockbox main menu.
06:53:23unstableo, I'm on the crappy of. I have a e280 v2. :(
06:53:33unstableCan rockbox work on the v2 yet? or the clip?
06:53:37saratogaunstable: no
06:53:53saratogaalso, the original firmware is offtopic here
06:54:03Unhelpfulnot really. you shouldn't install it if you're not testing or developing.
06:54:41 Join gromit` [0] (
06:55:13 Join AndyI [0] (i=AndyI@
06:56:52Unhelpfulsaratoga: i don't see any models in DeviceChart that claim to be ARM4... but i see some that are probably ARM and don't say which version?
06:57:28saratogaUnhelpful: armv4 is the ISA that ARM7 and some ARM9 cores run
06:57:38 Join Tuplanolla [0] (n=jani@unaffiliated/tuplanolla)
06:58:10Unhelpfulahhh... i see. core "versions" that don't sync with ISA versions, very nice :/
06:58:22 Join XavierGr [0] (n=xavier@rockbox/staff/XavierGr)
06:58:32saratogaPP, AMS, the Samsung chip in the GB f/X are ARMv4
06:59:54Unhelpfuli see... and only have 8-bit multiply? i should probably get somebody to time the two scaler versions on those devices, before i make assumptions about which is fastest...
07:00:23saratogatelechips can be either 4 or 5, and the beast is v6
07:00:50 Quit gromit`` (Read error: 110 (Connection timed out))
07:01:03saratogai don't think you have to worry about v>4 since those will be so fast anyway
07:01:52Unhelpfulwait a sec... so i've actually timed it on armv4, haven't i? when i tested the e200?
07:02:00saratogai think all modern v4 ARM cores have the 8 bit multiplier unit
07:02:05saratogayes PP is armv4
07:02:15 Join rocko_ [0] (
07:02:17saratogapretty much everything is v4 except the beast and some upcoming targets
07:03:19Unhelpfulthat makes me wonder if it will really be any faster on coldfire or sh-1 without the 64-bit multiplies
07:06:05 Quit _lifeless (Read error: 110 (Connection timed out))
07:09:01 Part toffe82
07:11:03saratogaUnhelpful: I noticed in your above patch that you are doing your fixed point multiplies in c
07:11:18saratogayou might consider stealing an assembly version from any of the codecs
07:11:42saratogathe speed up for codecs in using an assemblerized version of A*B >>C is typically enormous
07:11:50 Quit tajlero (Read error: 110 (Connection timed out))
07:14:26saratogaUnhelpful: see;revision=18570
07:16:22 Join Bagderr [241] (n=daniel@rockbox/developer/bagder)
07:16:46 Nick Bagderr is now known as B4gder (n=daniel@rockbox/developer/bagder)
07:19:25Unhelpfulwell, gcc is generating this, when compiling for e200:
07:20:22Unhelpfullooks like load, add the rounding value, umull the scaling value
07:25:37 Join cheesecake [0] (n=45c4bebd@gateway/web/cgi-irc/
07:25:51 Nick cheesecake is now known as love (n=45c4bebd@gateway/web/cgi-irc/
07:26:05 Nick love is now known as bunny (n=45c4bebd@gateway/web/cgi-irc/
07:26:27 Nick bunny is now known as keepsake (n=45c4bebd@gateway/web/cgi-irc/
07:27:04 Nick keepsake is now known as JackZhou (n=45c4bebd@gateway/web/cgi-irc/
07:27:59 Quit Horschti ("Verlassend")
07:31:06 Quit saratoga ("CGI:IRC (EOF)")
07:33:09 Quit Aurix_Lexico (Read error: 110 (Connection timed out))
07:33:14amiconnUnhelpful: Iirc a 32 bit division on SH1 would be faster than any multiplication involving 64 bits on it (using the libgcc routines)
07:35:00amiconnThis is because the division is done in asm, using SH1's division support instructions. This takes 64 cycles + some pre/postadjustment + function call overhead. The 64 bit multiplication stuff is plain C.
07:35:07Unhelpful amiconn: the SC_SHIFT = 24 case has enough overhead that one can expect the values not to overflow when multiplying by 63... so using /255 there is fine
07:35:25amiconnThe worst is divisions involving 64 bit, of course
07:36:25*amiconn wrote a 32*32->64 bit multiplication (signed, but that can be changed) in asm for SH1 and coldfire
07:36:41amiconn...for mandelbrot
07:37:16 Quit sbhsu (Read error: 110 (Connection timed out))
07:37:37 Quit AndyI ()
07:37:47amiconnActually, if the <<24 costs too much precision for larger images, it probably makes sense to introduce some of this asm rather than loose the precision
07:37:58amiconn>>24 of course
07:38:32amiconnSo basically you need 32*32->upper32 multiplication?
07:39:43Unhelpfulas it's written now, yes. the >>24 version gets by without. i can't say precisely where it the reduced precision starts to be noticeable, but white is still white up to at least 250K pixels input in the downscaling case
07:42:37Unhelpfulthe kind of graphics work i've done in the past was mostly with mmx... which is generally beastly, but has some vector NxN->topN instructions that are just too conveniently abused for fixed-point stuff
07:44:03amiconnarmv6 has a special instruction for this, btw, but only for signed
07:47:34 Join AndyI [0] (i=AndyI@
07:50:03***Saving seen data "./dancer.seen"
07:54:32 Join gerald [0] (
07:56:05geraldhi, I recently installed rockbox 3.1 on my first gen nano, and when I plug it into a car charger/fm transmitter it just keeps rebooting (maybe trying to enter usb disk mode?), any easy way to prevent that?
07:56:19geraldI have seen refs to some kind of car charger mode but that doesn't seem to be in 3.1
07:56:54cool_walking_Try holding MENU while connecting it.
07:59:54geraldawesome! thanks!
08:01:35cool_walking_You're welcome.
08:02:35 Quit GodEater_ (" ajax IRC Client")
08:02:51 Join GodEater_ [0] (i=c2cbc962@rockbox/staff/GodEater)
08:17:10 Nick Barahir_ is now known as Barahir (
08:17:46 Quit os2user (Remote closed the connection)
08:18:55 Quit BigBambi (Remote closed the connection)
08:27:19 Quit GodEater_ (" ajax IRC Client")
08:29:19 Join ender` [0] (
08:30:27advcomp2019i just back from shopping today and got a fuze and it is a v2 fuze
08:30:49linuxstbDoes that mean sandisk now have two firmware versions for the fuze?
08:31:20 Quit JackZhou ("CGI:IRC")
08:32:09*linuxstb spots it on Bagder's page
08:36:07 Quit gerald ("ciao")
08:37:19 Quit jhulst (Remote closed the connection)
08:37:55 Join Rob2222 [0] (
08:40:01 Join GodEater_ [0] (i=c2cbc962@rockbox/staff/GodEater)
08:42:18 Join kadoban [0] (
08:44:20 Quit cool_walking_ (" ajax IRC Client")
08:44:57 Part kadoban ("Konversation terminated!")
08:46:04 Join kadoban [0] (
08:49:10lucentI got bit by an ext4 corruption bug and lost my OS
08:49:17lucentjust getting back into things now
08:49:38lucentsorry I didn't test the non-HC microSD card on my Fuze and report back
08:50:06lucentadvcomp2019: I'm real interested to know what colour is your fuze?
08:50:44advcomp2019it is only a 2gb.. so it is only black
08:51:07lucentoh, 2gb and it's version 2 fuze, this is interesting. I don't know anything else about them
08:52:06 Quit tvelocity (No route to host)
08:54:20 Join tvelocity [0] (
08:54:29 Nick pixelma_ is now known as pixelma (n=pixelma@rockbox/staff/pixelma)
08:55:58 Quit Rob2223 (Read error: 110 (Connection timed out))
08:59:01 Quit Seed ("cu, Andre")
09:03:34 Join gregzx [0] (
09:18:05 Join petur [50] (n=petur@rockbox/developer/petur)
09:19:09 Quit gregzx (Read error: 104 (Connection reset by peer))
09:22:54 Quit sarixe ("Connection reset by the motherfucking peer")
09:22:57 Join os2user [0] (n=chatzill@
09:26:33 Join nibbler [0] (
09:26:58 Join Zagor [0] (n=bjorn@rockbox/developer/Zagor)
09:35:14 Join LinusN [0] (n=linus@gateway/web/cgi-irc/
09:40:18advcomp2019from what i have seen on the outsides of this v2 fuze and the photos of the v1 fuze, there is no way to tell if you get a v1 or v2 fuze
09:43:25 Part LinusN
09:43:27 Join LinusN [0] (n=linus@rockbox/developer/LinusN)
09:46:28 Quit yhuang (Read error: 60 (Operation timed out))
09:46:51B4gderso even the logos are exactly the same?
09:47:04B4gderafaiu they differ on the v1 vs v2 clips
09:47:14lucentB4gder: will wait for advcomp2019 to respond, but it won't surprise me
09:47:19Zagoryes, the roundness of the 's' is different
09:47:25Zagor(on clip)
09:47:37B4gderlucent: why? why wouldn't the fuze be like the clip in this aspect?
09:47:58lucentB4gder: the official line from SanDisk is to operate the device and check the firmware version installed
09:48:16B4gderyes, but I bet that too is the same for the clip
09:48:19lucentwhy re-tool just for a hardware revision?
09:48:33B4gderdid did it for the clip
09:48:39B4gderthey did
09:48:50advcomp2019B4gder, the sansa is the same as my v1 clip
09:48:56 Join Thundercloud [0] (
09:49:13B4gderlucent: you just repeat things as if the v1 and v2 clip difference doesn't exist
09:49:24B4gderadvcomp2019: ok
09:49:49 Quit nibbler (Read error: 110 (Connection timed out))
09:50:07***Saving seen data "./dancer.seen"
09:50:23advcomp2019so the best bet with the fuze is the firmware version still
09:50:38lucentB4gder: I'm just asking, you know, why would you re-tool for that
09:50:53B4gderand again, I repeat that THEY ALREADY DID FOR THE CLIP
09:50:54lucentmaybe it saves the mfg money or they have a supply problem
09:51:02lucentI don't know why they did it for the clip
09:51:13lucentgenerally you don't want to do that
09:51:18B4gderclearly they changed the logo
09:51:34Zagorlogos even
09:51:45B4gderso they probably want their newer devices to use their up-to-date logo
09:51:54B4gderI don't see what that is strange
09:52:13lucentthat doesn't have to be related to firmware version
09:52:20*B4gder sighs
09:57:26 Quit daurnimator ("Cyas later...")
10:05:04 Quit z35 ("Leaving")
10:10:49 Quit FOAD (Read error: 110 (Connection timed out))
10:18:32advcomp2019i just got out the broken 4gb fuze that someone gave me, it looks the same as this v2 on the outside too
10:19:33Zagoradvcomp2019: what is the flash size of your v2 fuze?
10:19:40 Join tyfoo [0] (
10:20:13 Quit tyfoo (Success)
10:20:18advcomp2019Zagor, i should look at that.. give me a few
10:23:47advcomp2019the "t" firmware says 8MB on the SDRAM test
10:24:13Zagorok, but I asked about flash :-)
10:25:24advcomp2019i am not sure what you mean then.. sorry
10:26:48Zagorwhat is the storage size? 2GB, 4GB etc?
10:27:08Zagorthe "disk"
10:27:09advcomp2019o ok.. 2gb fuze
10:29:16advcomp2019i got it at NFM(a store in omaha ne) at a 10% discount since office depot had them for $39.99
10:31:40 Join Anges [0] (
10:33:00lucentHi agaffney
10:33:03 Quit Thundercloud (Remote closed the connection)
10:33:03*lucent fails
10:33:11lucenthi Anges
10:33:43Angesoh, my keyboard was wrong, I said hi
10:34:46AngesI haven't tested the new rockbox yet, but I have bought a micro sd card
10:36:26lucentthere is code now which is more reliable for Fuze 8gb and a card
10:37:17lucentone developer fixed a bug with "Boost" feature, and suddenly I found the card works more easily
10:37:28lucentsomething evil is still lurking in the code
10:38:19Zagorsubtle fw differences. fuze1 uses e59ff048 for first vectors, while fuze2 has *58. and the chunk fuze1 has at offset 0x500 is on 0x510 for fuze2.
10:39:04Zagorthe double header seems to have different fields too. at least it's not the same amount of 0xff padding
10:39:22Zagorthough I don't remember what the specific fields were
10:40:19 Quit linuxstb (Read error: 110 (Connection timed out))
10:40:49 Join EspeonEefi [0] (i=eefi@SAFFRONCITY.MIT.EDU)
10:40:58advcomp2019Zagor, i think funman said the v2 clip and the v2 fuze was very similar but i could be remembering wrong
10:41:12Zagoryeah I would assume they are
10:41:20ZagorI'm just peeking a bit
10:44:47 Join Casainho [0] (
10:44:54Casainhohello :-)
10:45:45Casainhodoes anyone knows if there is some target player that uses SPI SD Card communication? - I am looking for SPI drivers for SD card... and not SD BUS card drivers...
10:47:11 Quit rocko_ ("Leaving")
10:47:14lucentCasainho: not related to rockbox but there's drivers which do this off an LED light signal
10:47:23advcomp2019Zagor, i might go and pick up another to see if i get lucky again in getting a v2 fuze so i can try some code out
10:48:20Casainholucent: thanks - I will then look for that drivers. I saw a lot of drivers for 8 bit MCU as AVR, etc... however I am having problems doing software reset to the card :-(
10:51:27amiconnCasainho: Ondio does SPI for MMC
10:51:39Casainholucent: :-) −− there is a lot of pages talking about that SPI drivers ;-)
10:51:54Casainhoamiconn: :-) :-) Thanks!! :-)
10:52:28lucentCasainho: I hope what I said is helpful, I'm not interested enough in these things to learn more
10:52:28amiconnThere are some gotchas in that driver though. Most important one is the nasty bitswap
10:52:52Casainhoamiconn: did you done that drivers?
10:52:59Casainholucent: thanks ;-)
10:54:40Casainhoamiconn: where in source can I find the Ondio? arm/ ??
10:55:54Angeslucent: I've « upgraded » rockbox, inserted my SD card, and got « *PANIC* // transfer error : -88 »
10:55:58lucentfind firmware/targets -iname '*ondio*'
10:55:59 Join DerDome [0] (
10:56:05Angeshow do I turn it off?
10:56:13lucentAnges: hold power for many seconds
10:56:13 Join n1s [0] (n=nils@rockbox/developer/n1s)
10:56:17lucentmaybe 12+ seconds
10:56:21Angesgot it
10:57:00Angesnow I see the SD card and can read ogg files on it
10:57:12Casainholucent: oh, I am on a windows box :-( −− at work...
10:57:15lucentreally? that's a nice thing to hear
10:57:28*lucent points out that IRC lag can be ironic
10:57:40Angesthe rockbox logo is a little weird though
10:57:48lucentAnges: dots?
10:57:58Casainholucent: oh, I used Eclipse search and I found it ;-)
10:58:08Angesdots on the beginning screen
10:58:14*lucent <3 eclipse
10:58:23lucentAnges: I see those dots also
10:58:34lucentit's just for 8gb fuze, no other player
10:58:37Angesand some euh… something else when I'm playing music
10:59:02lucentthere is some problem that happens on all fuze players in WPS (the "playing music" screen)
10:59:11lucentsome discoloured bar of pixels
10:59:52AngesI can't hear anything, but the fuze still pretends it's reading music
11:00:03lucentoh, maybe decoding is not working
11:00:04Casainholucent: on sh/archos/ondio, I can't find any driver for SPI SD card... just a driver for ATA(?)
11:00:16ZagorAnges: the fuze is in very early stages. it's not supposed to be working.
11:00:20Angesno, it's stuch
11:00:33AngesZagor: I know :)
11:00:50lucentAnges: some FLAC format audio will play for me
11:01:02lucentogg needs "boost" which isn't working fully, mp3 causes reboot
11:01:05Angesat least, now SD card is recognized (it's a 2 gb)
11:01:17lucentthere are some ugly bugs that need to be found
11:01:25lucentthat's a good report Anges, thank you
11:02:18 Quit midkay_ ("Leaving")
11:02:58AngesI tried again, and it's stuck when: turn on fuze, start play music on sd card, go back to explore, select a file on fuze instead of sd card
11:03:12Casainhoamiconn: I can't find the SD card SPI drivers under sh/archos/ondio....
11:03:25Anges(unstuck with pressing play twice)
11:04:41Angeswhat's really weird is that grey logos are fuzzy, and yellow ones are ok
11:08:54lucentAnges: the graphic problems are known, but the bug that causes those problems is not known
11:09:04lucentwe have a work-around that hides the graphic errors
11:09:33AngesI wish I could help more
11:09:37lucentme too
11:10:21lucentsending a Sansa Clip 2GB or greater to an interested developer, this is a good way to contribute
11:11:04lucent2GB and larger Clip has the special storage device, which our Fuze players have and makes files over 1GB address unreadable
11:11:27lucentsupport for this Bank Switching feature is needed and hardware is a must
11:12:33lucentAnges: I heard IRC user fdinel (not presently online) wants a Clip to play with
11:12:52 Quit Casainho ("ChatZilla 0.9.84 [Firefox 3.0.5/2008120122]")
11:13:41tmztlucent: does it work if the partition is resized?
11:13:49Angesdo I get it back when he's finished playing with it?
11:13:54lucenttmzt: it does on my Fuze, yes
11:14:21Angestmzt: I tried unfragment instead of resizing
11:14:23lucenttmzt: I must warn you that what I just said is answering your question, but that whole exchange is faulty
11:14:42lucenttmzt: you must NOT change partitioning on the Fuze or it will fail
11:15:05lucentI run my Fuze with an 800mb filesystem, on an 8gb block device
11:15:29tmztwhat will fail and how did you change it then?
11:15:40lucenti.e. for explanatory purposes, you could say I have 7200mb unused data storage space
11:16:09lucentthe filesystem is only so big and then the rest is unused and avoided because the filesystem never goes there
11:16:15tmztoh, I see you did not change the partition table at all just dd'd a filesystem of that size?
11:16:26lucenttmzt: I used mkfs.vfat size option
11:16:58lucentit warns and refuses, and I use the force option in superfloppy mode, and it succeeds
11:17:14lucentsomething like that, yes
11:38:26 Quit EspeonEefi ("さよなら")
11:50:09***Saving seen data "./dancer.seen"
11:55:08 Join linuxstb [0] (n=linuxstb@rockbox/developer/linuxstb)
12:07:30 Join tyfoo [0] (
12:24:21 Join reacocar1 [0] (i=reacocar@
12:30:58 Quit reacocard (Read error: 110 (Connection timed out))
12:35:51 Quit gevaerts (Nick collision from services.)
12:35:59 Join gevaerts [0] (n=fg@rockbox/developer/gevaerts)
12:38:40 Join culture [0] (
12:42:03 Quit BHSPitLappy ("Ex-Chat")
12:47:16 Nick fxb__ is now known as fxb (
12:51:18 Join MTee [0] (n=mtee@
12:56:21 Join TheSphinX^ [0] (
13:08:24*B4gder suprises everyone with a mail!
13:08:38B4gderbut my spelling is still the same crappy one
13:09:36 Quit japc_ (Read error: 110 (Connection timed out))
13:12:42*scorche wonders what is surprising about this ;)
13:13:51 Join tajlero [0] (
13:14:27*n1s would really like to see plugin localization getting finished
13:16:49 Quit feisar ("leaving")
13:17:09 Join feisar [0] (
13:19:46 Join fredddy [0] (
13:20:15 Quit TheSphinX^ ("XChat@Linux")
13:22:24 Nick JdGordon|afk is now known as JdGordon (n=jonno@rockbox/developer/JdGordon)
13:22:32 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
13:32:51*JdGordon would really really really like some hwcodec testing done on 9795....
13:33:14JdGordonotherwise its going in once its working on swcodec and someone else can fix any possible hwcodec bugs...
13:33:43 Quit __lifeless (Remote closed the connection)
13:34:00 Join __lifeless [0] (n=lifeless@
13:43:51 Join abholschein_ [0] (
13:50:12***Saving seen data "./dancer.seen"
13:51:02 Nick JdGordon is now known as JdGordon|zzz (n=jonno@rockbox/developer/JdGordon)
13:55:26 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
13:56:56 Quit abholschein_ ("Ex-Chat")
13:57:53 Quit tim__ (Connection reset by peer)
13:57:56 Join timc`` [0] (n=aoeu@
13:57:56 Join kachna|lappy [0] (
13:57:58 Join n00bi3 [0] (
13:58:00 Quit n00bi3 (Read error: 104 (Connection reset by peer))
14:05:50 Join tyfoo2 [0] (
14:06:10 Quit tyfoo2 (Read error: 104 (Connection reset by peer))
14:07:53 Quit kachna (Read error: 113 (No route to host))
14:11:00 Join PaulJam [0] (
14:14:24 Join tyfoo2 [0] (
14:17:53 Quit agaffney (Read error: 110 (Connection timed out))
14:19:49 Quit tyfoo (No route to host)
14:19:49 Nick tyfoo2 is now known as tyfoo (
14:20:18 Join agaffney [0] (n=agaffney@gentoo/developer/agaffney)
14:27:42 Join Carz [0] (
14:30:09 Quit tajlero ("Leaving")
14:38:42kadobandoes anyone have experience profiling the simulator (gprof)? i have it almost working, but all of the times in the "flat" output or whatever it is are zero
14:38:47kadobanwhich kind of defeats the purpose
14:40:16kadobanthe only thing that i can guess is that it's because the simulator never exits on its own so i have to kill it. does the prevent the final bit of information being output to the file or something? (is there a way to make the simulator exit more cleanly?)
14:41:11 Join Aurix_Lexico [0] (
14:44:52 Part B4gder
14:47:07 Join robin0800 [0] (
14:48:56CarzI need to find out when my ipods warranty runs out... sometime soon, I think\
14:55:27GodEater_Carz: why would you care ?
14:55:27 Quit os2user (Remote closed the connection)
15:01:09Unhelpfulkadoban: write a plugin that calls exit?
15:02:53 Join Acky [0] (
15:04:09kadobanUnhelpful: hmm...that might be an idea. i'll give it a try. unfortuantely i'm starting to think that that's not my problem, but we'll see...
15:04:58 Nick fxb is now known as fxb__ (
15:07:08 Quit Acksaw (Read error: 60 (Operation timed out))
15:09:22Unhelpfulit's easy to try... assuming that the sim plugin loader won't bar that from working
15:09:35 Join japc_ [0] (n=japc@
15:11:01CarzWhy wouldn't I care?
15:12:44kadobanUnhelpful: yeah, it worked like a charm, except gprof still isn't giving me any times. it doesn't mention anything about functins from my plugin either so i think maybe it's a build thing (plugin not getting build with right cc, ld flags?)
15:15:01Unhelpfulweird... it might be, or it might be a dlopen thing
15:15:48kadobani'm pretty awful at figuring out build stuff, i think i'll just try to get sysprof working
15:16:28Unhelpfulamiconn: i'm pretty sure i found the routines you meant... the arm one is pretty clearly just a smull and a shift :)
15:17:01Unhelpfulkadoban: just don't complain when you have trouble driving nails with your sledgehammer
15:17:39kadobanUnhelpful: don't worry, my sledgehammer is apparently broken too. lucky me
15:34:33 Join t0mas [0] (n=Tomas@rockbox/developer/t0mas)
15:44:53 Join TheSphinX^ [0] (
15:50:17***Saving seen data "./dancer.seen"
15:56:40kadobanis there a way to get our make to tell me what exact commands it is running? not sure if i have the right CFLAGS being set (i don't see anything general in the make manpage)
15:58:04kadobanscratch that, i'm blind
16:00:15Zagorkadoban: parameter V=1 does that
16:00:45*Unhelpful wonders if dlopen interferes with profiling at all
16:01:26kadobani'm trying the actual "source profiling" thing described in docs/TECH now, as soon as i figure it out. i'm hoping that works
16:01:49 Join Horscht [0] (n=Horscht@xbmc/user/horscht)
16:11:52 Quit xSlack (Read error: 60 (Operation timed out))
16:12:29 Join xSlack [0] (
16:15:03 Join archivator [0] (i=foobar@
16:15:31 Nick archivator is now known as Guest73098 (i=foobar@
16:16:43 Nick Guest73098 is now known as archivator (i=foobar@
16:22:02kadobanso i'm trying to get profiling working. i configured with profiling support, added $(PROFILE_OPTS) to the cflags in apps/plugins/plugins.make, and surrounded the hdr->entry_point call in apps/plugin.c with profile_thread() and profstop(), but i get an "Undefined instruction at 0x40004B46 (0)" error when i run my plugin. could anyone suggest something i might be doing wrong?
16:26:03 Join tyfoo2 [0] (
16:27:00*dionoea thinks thtat kadoban should have tried oprofile first
16:29:05kadobandionoea: hmm, i'll look into that if this next test doesn't work
16:29:33dionoeathe only drawback is that it's a kernel level statistical profiler (you need to run linux)
16:30:07kadobanoh good, "data abort" this time, how nice
16:30:15kadobanyeah, i'm in linux
16:30:42dionoeathe main advantage is that you don't need to do anything to the code (just compile with debug symbols)
16:30:54dionoeaand it also has no impact on performance
16:33:01 Quit t0mas ("bye!")
16:37:06 Join flydutch [0] (
16:37:50 Join Davekong [0] (n=Davekong@
16:38:05 Part Carz
16:38:07 Join gartral [0] (
16:39:45gartralumm.. i noticed that the latest build says it has working USB, but i cant figure out how to access it.. the little instruction in the changelog isnt too clear
16:40:21 Join MethoS [0] (
16:40:32 Quit tyfoo (Connection timed out)
16:40:32 Nick tyfoo2 is now known as tyfoo (
16:40:49gevaertsgartral: where does it say so?
16:41:21gartralaorn it says so in the second entry down in the changelog
16:42:16*Zagor fails to translate that commit message to "USB is now working"
16:42:48rasherAny thought on FS #9666? I can think of a few situations where inverting the display might make sense, but none where you'd want albumart to get inverted also
16:43:32gartralthe part that say the "detect USB connection by bus reset. When received, disconnect and restart the driver fully enabled."
16:44:15kadoban...if USB was enabled by default suddenly, there would be much more warning
16:44:24gevaertsThat's the low level driver
16:44:28kadobanor notification i should say
16:44:43gevaertsIndeed. There would be beer
16:44:46gartralahh.. like i said, it wasnt too clear
16:46:22linuxstbgartral: The start of that commit message is the important bit - "USB detection changes"...
16:46:26 Join toffe82 [0] (
16:50:17ajblogbot: last Unhelpful
16:50:25 Quit mrkiko (Remote closed the connection)
16:50:31Unhelpfuli'm here. :)
16:50:38ajbAhh, so you are
16:50:59 Quit robin0800 (Read error: 104 (Connection reset by peer))
16:52:02 Quit Zagor ("Client exiting")
16:53:38ajbUnhelpful: Can you commit and close FS #9758 - I've just fixed the commentry on your version of the patch.
16:53:56Unhelpfulhrm, yes, it looks like on sh and coldfire, (uint32_t)a * (uint64_t)b >> 32 uses __muldi3 :/
16:54:24Unhelpfulajb: let me give it a look first, but i think we've answered any objections that were raised, at this point.
16:54:25 Join sarixe [0] (n=sarixe@
16:54:48 Join arohtar [0] (
16:55:16ajbUnhelpful: There is milage in putting more info in rockbox-info but this at least stops the timeout for non git-svn repos which was the original problem I had :-)
16:56:00 Join nplus [0] (n=nplus@243.131.Globcom.Net)
16:56:02Unhelpfulindeed... pure git repos are useless until *something* gets in to skip using git-svn
16:56:59*ajb returns to cleaning up his playlist patch
16:57:31Unhelpfulwell, i'd move the "is this a git-svn commit" comment down a couple lines, and put a "get revision of latest git-svn commit" there instead... otherwise, looks about done :)
17:01:18ajbUnhelpful: v6 uploaded
17:01:35ajbUnhelpful: hang fire wrong FS
17:01:54Unhelpfulit's ok, i'm making the comment change anyway, about to commit. :)
17:01:58 Quit arohtar (Client Quit)
17:02:02ajbAhh ok
17:02:05ajbUnhelpful: thanks
17:02:16 Quit agaffney (
17:02:22 Join faemir [0] (
17:02:23NJoinagaffney [0] (n=agaffney@gentoo/developer/agaffney)
17:02:53ajbDoes git:// get automatically updated when SVN is tweaked or is it a cron thing?
17:03:32 Join {phoenix} [0] (
17:04:20UnhelpfulI *think* it's automatic, but i believe Zagor would be the one to ask to find out for certain.
17:05:27rasherOh by the way, a good reason not to switch to git (yet): Windows support is still pretty poor.
17:06:21 Quit tmzt (Read error: 145 (Connection timed out))
17:07:28Unhelpfulrasher: i'm also not sure we'd truly gain anything by making the "central" repo use a distributed system
17:07:57Unhelpfulother than headaches, and some changes in branch/merge capabilities
17:08:32ajbrasher: I don't think anyone has suggested it recently. I think it's fine as long as we try not to break other VCS's by hardwiring too much to SVN
17:09:11 Join tmzt [0] (
17:09:40ajbrasher: I only use git locally so I can do many experimental local commits not having commit privilages to the project. DVCS's do make it easier for non-commiti priv devs to hack about
17:10:54GodEater_Unhelpful: there's a post-commit hook in svn that updates the git repo whenever someone commits code.
17:11:05GodEater_so it's automatic :)
17:11:06UnhelpfulGodEater_: that's what i had thought :)
17:15:51 Join yhuang [0] (
17:19:13 Join MethoS- [0] (
17:21:19 Join gustavwag [0] (n=c9dfcc73@gateway/web/cgi-irc/
17:21:28ajbwrt to SoC suggestions I think getting Qemu support for a few of our embedded targets offers some oppertunities
17:21:29 Quit gustavwag (Client Quit)
17:21:46 Join preglow [0] (
17:24:32GodEater_surely that's a Qemu project rather than a rockbox one though ?
17:25:50ajbGodEater_: well more of a cross project. Afterall the emulation of the device would have to be very driven by what Rockbox knows about them. But I take your point
17:29:22ajbohhh, I like the improved simulator - saves me keeping the keys wiki open!
17:30:11gevaertsThat's been there for a long time...
17:30:14linuxstbajb: You may have found the "−−background" option useful in the past then...
17:30:17 Part Davekong ("Leaving")
17:30:37*ajb wonders why he's never seen it before
17:32:36 Quit MethoS (Read error: 110 (Connection timed out))
17:33:24n1skadoban: I had a problem with the rockbox source profiling when trying it a while ago connected to it keeping it's own copy of the api pointer that had to be initialized, not sure how this was affected by Unhelpful's plugin api changes though
17:35:08kadobann1s: oh, hmm. i didn't see anything like that, i'll have to look into it. i'm using oprofile now, but i think the differences between target and sim are too great for it to be really useful
17:35:50 Quit nplus (Remote closed the connection)
17:37:17 Part unstable
17:40:13 Join gregzx [0] (
17:45:41Unhelpfuln1s: profile_plugin doesn't do anything *to* the api pointer, though... it just uses the global one now
17:46:14kadoban:( i was hoping i was missing something. then i have no idea why it's not working
17:46:24n1sUnhelpful: ah
17:46:47 Join shyam_k [0] (n=user@
17:48:16shyam_kis there any kind of voice command tools for rockbox? is it actually imaginable? these days, cellphones too are having voice enabled dialers and all which runs on small but fast recognizers for limited vocabulary..
17:48:46dionoeano such thing exists as far as i know (for rockbox)
17:48:55 Join saratoga [0] (n=9803c6dd@gateway/web/cgi-irc/
17:49:36rashershyam_k: cellphones are quite powerful compared to the hardware rockbox runs on
17:49:42shyam_khmm yeah
17:49:59shyam_ki had done a project trying asr engine sphinx on ngw100
17:50:06shyam_kand it worked to my surprise..
17:50:19***Saving seen data "./dancer.seen"
17:52:49Unhelpfulthe function names in there make me wonder if profile_plugin is cygwin-specific?
17:53:14 Quit TheSkunkMan ("Leaving")
17:53:48shyam_kbut what would be rockbox's idea about such things?
17:54:05shyam_kmay be a customized small version? or its simply out of scope?
17:54:32kadobanUnhelpful: that would be really annoying. the only documentation i can find mentions nothing about that, but that doesn't mean much, 'cause it's pretty sparse
17:55:23n1sUnhelpful: no, it works on target
17:55:26ajbHave there been any attempt to redo the virtual keyboard?
17:59:03 Quit petur ("work->road-shooting->home")
18:00:13 Quit jfc (Read error: 104 (Connection reset by peer))
18:01:29shyam_kwhich would be the best platform to work with rockbox? to play to the fullest?
18:01:51*shyam_k is yet to have his portable media player..
18:02:20PaulJamajb: what's wrong with the current implementation if the virtual keyboard?
18:02:21 Join MethoS-- [0] (
18:02:30 Quit {phoenix} (Remote closed the connection)
18:03:31linuxstbshyam_k: The answer depends on what you want - there is no single "best" device that runs Rockbox -
18:03:32ajbPaulJam: It's complete but slow. The ordering of letters and the fact they are all there leads to excessive clicking to do a simple word.
18:03:33saratogais ptw419 or any of the original Gigabeat S hackers ever online?
18:03:45 Join domonoky [0] (n=Domonoky@rockbox/developer/domonoky)
18:04:02toffe82saratoga: there are not there anymore
18:04:06kadobani /think/ you can change the layout of the letters relatively easily
18:04:09PaulJamajb: you can customise the vkeyboard.
18:04:24ajbPaulJam: mainly it's a feeling there could be a more intuitive way of doing it, especially on the bigger screen devices
18:04:36ajbPaulJam: is this settings magic?
18:04:38toffe82sarwhat do you need ?
18:04:44saratogatoffe82: who should I talk to about hacking the Gigabeat T then?
18:05:20shyam_klinuxstb: as i said, i wanna have the one with most computational power:)
18:05:39saratogathen you want a gigabeat S
18:05:58shyam_ksaratoga: to me?
18:06:11toffe82saratoga: it is aibotpet who found the hole in the security
18:06:35toffe82aibopet ^^
18:07:09*n1s thought he was called "zunepet" ?
18:07:16linuxstbDid MarcGuay ever solve his Beast problem?
18:07:39 Quit PaulJam (".")
18:07:40toffe82saratoga: the problem for the T is that there is no updater and we don't have a copy of the firmware, we have the one for the t401 but not the t400
18:07:44 Join PaulJam [0] (
18:08:02toffe82zunepet aibopet are the same person
18:08:02ajbPaulJam: now I didn't know you could do that :-) thanks
18:08:12saratogatoffe82: if the CPU is the same, could I flash and recovery like i can on the S?
18:08:27saratogasorry flash and recovery the nk.bin
18:08:40saratogaof course if the USB hardware isn't precisely the same that could end badly
18:08:43toffe82saratoga: the electronic is the same as the S
18:08:51saratogais it possible to recover a T you think?
18:08:59toffe82from what I saw when I open it
18:09:10toffe82check the picture on the wiki
18:09:20saratogathe hardware is the same but that doesn't mean all the GPIO pins are wired the same
18:09:37saratogaevery single chip on the fuze and clip is the same, but the wiring is a bit different for instance
18:09:44jhMikeSI'd do some investigating and disable alot of stuff before assuming that. It could do nasty things if some key things differ.
18:10:14saratogais it possible to unbrick the T?
18:10:35toffe82no original firmware available
18:10:44toffe82what model do you have ?
18:10:49toffe82t400 or t401 ?
18:10:55saratogathe non wifi one
18:11:11saratogayes 400
18:11:14toffe82the one I have too
18:11:17saratogasorry didn't realize that was on the back
18:11:38*ajb thinks the standard ASCII layout would probably make a sensible default keyboard layout
18:11:41saratogaif I could get the OF would it be possible to unbrick you think?
18:12:36toffe82I think it should work like the S
18:13:17 Join PaulJam_ [0] (
18:13:36toffe82and toshiba is very conservative, the updater as the same function for the S, V and T so the recovery mode is the same
18:14:36toffe82the only thing is to be sure you can get the firmware
18:15:37 Part gartral
18:17:35 Quit MethoS- (Read error: 110 (Connection timed out))
18:17:41 Join jgarvey [0] (
18:21:56 Quit gregzx ("ChatZilla 0.9.84 [Firefox 3.0.5/2008120122]")
18:24:52 Join {phoenix} [0] (
18:26:10 Join tim__ [0] (n=aoeu@
18:26:52ajbI seem to be able to crash the simulator by holding down a scrollwheel button
18:29:19ajband now I can't... odd
18:31:31 Part LinusN
18:33:32gevaertsjhMikeS: did you see FS #9812 ?
18:33:39 Join jfc [0] (
18:40:10 Quit PaulJam (Connection timed out)
18:40:31 Join kugel [0] (n=chatzill@unaffiliated/kugel)
18:41:01shyam_know the question: is there an "official" rockbox project to develop a good clean, free/open hardware player?
18:42:26kugelthere's such a project, yes
18:42:53gevaertsHow official do you want it?
18:42:58kugelofficial or not is debatable, but it aims for a free and open player which runs rockbox
18:45:05gevaertsI personally wouldn't consider it official at all. There's not a single rockbox committer involved
18:45:56shyam_kand in the forum i finds it.. on ngw100?!
18:46:14shyam_kin which i ran the sphinx successfully:)
18:46:28*gevaerts has no idea what shyam_k is talking about
18:47:31shyam_kor the one you said is some other?
18:47:36jhMikeSgevaerts: shouldn't do that. I thought those had distinct inputs for detecting USB/Charger
18:47:57gevaertsshyam_k: that's the only one I know about
18:48:28jhMikeSIf not, the framework is in place of course to deal with it.
18:48:50shyam_kgevaerts: like say openmoko people say about their phone, wouldn't it be nice if rockbox says about a general hardware which works well.. and may be some merchandise too!:)
18:48:52gevaertsjhMikeS: I'll test later today. Maybe it just needs to test whether power remains available after detecting a reset?
18:49:08shyam_kthe free software way!
18:49:40*ajb uploads final FS #9677 (Quick save menu instead of warning if a dynamic playlist is due to be erased)
18:49:57shyam_kwhy rockbox wanna sit behind the warranty threat of these bullshit companies!:)
18:50:45 Join MethoS [0] (
18:51:06jhMikeSgevaerts: ipod video does have distinct inputs according to the debug screen for usb/charger. Perhaps power code isn't doing it right.
18:51:24gevaertsshyam_k: feel free to work on such a thing, but don't forget that most or all rockbox devs are software people
18:51:44ajbCan anyone review FS #9677?
18:51:47 Quit MethoS-- (Read error: 60 (Operation timed out))
18:52:07jhMikeSOr the adapter there uses USB power regardless. Perhaps the arc driver should always use the bus reset method.
18:52:24gevaertsI think that would be best, yes
18:52:45ajbDoes rockbox have a unused parameter macro?
18:52:49 Quit timc`` (Connection timed out)
18:52:55shyam_ki just asked.. seems it would have already discussed here no?
18:53:23*shyam_k doing final year undergraduate course on electronics
18:53:25jhMikeSIt might allow the beast to use a USB-style cellphone charger too (but then it won't configure :\)
18:53:35shyam_kwould be more than happy to work on such a project too..
18:53:52gevaertsI guess the video has power inputs on its dock connector somewhere, but those won't be used if you use the USB cable to charge
18:54:23shyam_kand for years were thinking to develop my own ogg player.. was thinking it around vs1000 vlsi chip once.. but droped as i don't have tech to handle the chip in my vicinity
18:55:20gevaertsjhMikeS: one of the reports was e200
18:56:04 Quit linuxstb (Read error: 113 (No route to host))
18:59:00jhMikeSgevaerts: The reports were from yesterday before committing the other refinements. e200 should be solved.
18:59:32gevaertsjhMikeS: the last comment is from today
19:01:49 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
19:02:33 Join miepchen^schlaf [0] (
19:03:02jhMikeSmeh, that just sounds utterly whacked. Those shouldn't have any connection.
19:03:51 Join MethoS- [0] (
19:04:06 Join Seed [0] (
19:04:07jhMikeSI'll tape over the data lines and check again. WTF.
19:04:13 Join TheSkunkMan [0] (
19:06:06*jhMikeS wonders if the charger is out to get him by faking a bus reset somehow.
19:06:34jhMikeSif so then how would the backlight even matter?
19:06:40 Join JaK [0] (i=47cf9dac@gateway/web/ajax/
19:07:10gevaertsI have no idea about that. Backlight shouldn't be involved at all
19:07:25 Join nibbler [0] (
19:08:57 Quit MethoS (Read error: 60 (Operation timed out))
19:09:32 Join BigBambi [0] (n=Alex@rockbox/staff/BigBambi)
19:10:07TheSkunkManhey could someone help me with using the SimpleFirmwareCrack(decode_light.cpp). How do I input the firmware file?
19:10:38jhMikeShehe, I'll be a $%@!. It didn't do that before.
19:14:45 Join blkhawk- [0] (
19:15:48 Quit blkhawk (Read error: 60 (Operation timed out))
19:16:09JaKim trying to do a manual install
19:16:13JaKwhat does it mean by "root"
19:16:43 Nick blkhawk- is now known as blkhawk (
19:17:46ajbJaK: In context?
19:17:57JaKwell let me start from begining
19:18:17JaKwhen i try to install using the utility, it finds the sansa
19:18:21saratogatry google if you find computer words you're not familar with
19:18:24JaKbut i use complete install
19:18:26 Join Thundercloud [0] (
19:18:33JaKit says no sansa detected
19:18:42 Join karashata [0] (n=karashat@
19:19:01ajbJaK: Are you running rbutil as the super user (i.e. as root)?
19:19:14ajbJaK: Are you in fact running rbutil on a Unix?
19:19:46ajbJaK: Which manual are you reading?
19:20:05JaKoh one sec
19:20:09JaKlet me check sommit irl fast
19:20:21JaKnvm, kk im reading the sansa manual
19:20:22 Join Nico_P [50] (n=nicolas@rockbox/developer/NicoP)
19:20:29 Join blkhawk- [0] (
19:20:31JaKe200 series
19:21:15ajbJaK: this one?
19:21:44ajbWhich section?
19:21:55JaKim looking at section 2
19:22:10ajbJaK: You mean root as in "root of your player’s drive. "?
19:22:34ajbahh - I see where confusion can come in
19:23:30JaKbtw can u tell if im doing somthing wrong, i try to do the quickstart installation, and it says that no sansa detected when the location is alredy set up
19:23:38ajbThe root of the sansa filesystem would be where it appears on your machine. I'm guessing on Windows it would a drive letter? (not having used Windows for 7 years I have no idea where XP puts these things)/
19:23:56 Quit blkhawk- (Read error: 104 (Connection reset by peer))
19:23:56JaKoh ic
19:23:57jhMikeSgevearts: It seems it gets a phony bus reset indeed.
19:24:00JaKwell ive beend oing that right
19:24:48 Join blkhawk- [0] (
19:25:02JaKis there a way i can delete the out of the box software/
19:25:04gevaertsjhMikeS: maybe wait for a while after detecting a bus reset and see if you still have power?
19:25:15n1sJaK: if sansapatcher can't find the sansa you probably need admin privileges
19:25:21 Join mcuelenaere [0] (n=mcuelena@rockbox/developer/mcuelenaere)
19:25:36jhMikeSI thought about checking the B-session valid bit at the time it gets it.
19:25:44JaKbut i do have admin privlages
19:25:49mcuelenaeregevaerts: has the Rockbox USB storage driver been tested on WinVista?
19:26:12gevaertsmcuelenaere: I think so, yes. It's been a while though
19:26:13jhMikeSIf the 4V4 is no good, it's a bad reset.
19:26:20PaulJam_JaK: what is the version of the original firmware on your e200?
19:26:20n1sJaK: is the sansa in msc mode? (does it appear with a drive letter?)
19:26:37JaKit does
19:26:44gevaertsmcuelenaere: not by me anyway
19:26:53mcuelenaereand it worked?
19:27:17JaKPaulJam_: 1.01.18A
19:27:22gevaertsAs far as I remember, yes
19:27:49mcuelenaereok, then it's probably the driver
19:27:54 Quit blkhawk- (Read error: 104 (Connection reset by peer))
19:28:28 Join blkhawk- [0] (
19:28:57JaKi jsut need to install the bootloader
19:29:08mcuelenaeregevaerts: and again I'm having 203/62695104 as ACK/NACK ratio, is that normal?
19:29:09 Join MethoS-- [0] (
19:29:20JaKdunno how to do that
19:29:47gevaertsmcuelenaere: That's a bit high. Are you tracing from the same PC?
19:30:26mcuelenaeretracing from the same PC gives me unreadable results
19:30:27JaKwhy does it keep saying "no sansa detected"
19:30:31mcuelenaerewhich is kinda annoying
19:31:30JaKauto detect works
19:31:37 Join blkjawk[work] [0] (
19:31:49gevaertsIf you need to use the same PC, an add-in USB card should help
19:32:00mcuelenaeregevaerts: I'm on a laptop
19:32:30gevaertsThere are cardbus USB controllers :)
19:32:40mcuelenaereand it seems like all the 3 ports are on the same usb controller, unless if they use isonchronous endpoints :/
19:32:47mcuelenaeregevaerts: could be, but I don't have them :)
19:33:04gevaertsMultiple controllers would be unlikely
19:33:31PaulJam_JaK: you are using rbutil, right?
19:33:50gevaertsAnyway there's always a possibility that we handle something the wrong way. Can you compare with the OF or with another storage device?
19:34:03mcuelenaerethat was what I was going to do
19:34:15mcuelenaerenow I'm hoping the indexing will not crash :)
19:34:57 Quit blkhawk (Connection timed out)
19:35:05 Nick blkjawk[work] is now known as blkhawk (
19:35:10mcuelenaerehmm it seems like I'm transmitting invalid usb scsi packets
19:36:08mcuelenaeregevaerts: can I send you the .mqu file?
19:37:01PaulJam_JaK: there seeems to be a bug with the current version of rbutil causing the "No Sansa found" error. But i'm not sure what the recommended woraround is.
19:37:21JaKwop! bootloader installd
19:37:29JaKlol i r troubleshooter pro =D
19:37:46krazykitPaulJam_, one possible workaround would be to just do a manual install
19:38:14 Join FOAD [0] (
19:38:33 Join bertrik [0] (
19:38:38PaulJam_JaK: how did you solve it?
19:40:00JaKi manually installd bootloader
19:40:14PaulJam_ah ok.
19:42:16Unhelpfulamiconn: whenever you're available to bother... i'm trying to figure out your sh-1 multiplier. it would seem to me that, if we're allowed to do 16-bit left shifts on macl+mach as a single 64-bit unit, that perhaps the mac instruction would be suitable for doing an unsigned multiply?
19:42:19JaKwhere can i access fonts...?
19:44:05Unhelpfulusing your ab*cd notation, it would be something like clrmac, mac a*c, shift left 16, mac a*d, mac c*b, shift left 16, mac b*d, leaving the complete result in macl/mach?
19:44:38JaKnvm soz ajb
19:44:39JaKim retarded
19:44:44ajbJaK: Most of your questions should be answered by
19:44:47ajbJaK: np's
19:45:50 Quit MethoS- (Read error: 113 (No route to host))
19:45:53mcuelenaeregevaerts what exactly should be done when the usb storage driver does usb_drv_send(ep_in, NULL, 0); ?
19:46:54*ajb is done for the day
19:47:47 Quit blkhawk- (Read error: 110 (Connection timed out))
19:48:46 Quit JaK (" ajax IRC Client")
19:50:23***Saving seen data "./dancer.seen"
19:53:04 Quit faemir ("Leaving")
19:56:25 Quit japc_ (Read error: 145 (Connection timed out))
19:58:59 Quit kugel (Nick collision from services.)
19:59:02 Join kugel_ [0] (n=chatzill@
19:59:12 Nick kugel_ is now known as kugel (n=chatzill@
19:59:35 Join z35 [0] (
20:00:25 Join faemir [0] (
20:00:26jhMikeSgevaerts: A VBUS (OTGSC bit 9) valid seems to give a proper indication if it's a spurious reset (since it should be above 4V4)
20:01:02 Part faemir
20:01:20 Quit Seed ("cu, Andre")
20:20:30 Quit Nico_P (Remote closed the connection)
20:21:06gevaertsmcuelenaere: usb_drv_send(ep_in, NULL, 0); should send a zero-size transfer
20:21:24mcuelenaereyes, I fixed that now
20:21:38mcuelenaeremass storage is a lot faster when initing now
20:21:46 Join Lear [0] (n=chatzill@rockbox/developer/lear)
20:23:05mcuelenaeregevaerts: is it normal that the OS constantly sends test_unit_ready 0 & test_unit_ready 1 while the device replies with UMS_STATUS_GOOD?
20:26:21gevaertsmcuelenaere: not sure actually. Maybe it's polling to see if you eject. How often is this polling?
20:26:33gevaertsjhMikeS: that's good news :)
20:27:04mcuelenaerehmm I'm not getting it atm and the test-pc is unavailable currently
20:27:36 Join FOAD_ [0] (
20:27:41mcuelenaereit's about 750ms-1000ms on my pc (WinXP)
20:27:48bertrikmcuelenaere, I think windows polls about once per second
20:27:55mcuelenaereok then that's normal
20:31:26 Quit MethoS-- (Read error: 104 (Connection reset by peer))
20:32:13 Join MethoS [0] (
20:32:43jhMikeSgevaerts: I'm pretty sure windows polls to watch for ejects and inserts. It must poll pretty often since it's rather responsive to it.
20:33:29gevaertsOk. Two votes vor regular polling, none against. That means it polls :)
20:35:15jhMikeSNothing informs the host that this is happening so it would have to keep polling. At least that's what I thought.
20:35:45gevaertsIndeed. It's reasonable behaviour
20:40:36 Join linuxstb [0] (n=linuxstb@rockbox/developer/linuxstb)
20:40:54 Join kugel_ [0] (
20:42:07 Quit kugel ("ChatZilla 0.9.84-rdmsoft [XULRunner]")
20:42:19 Nick kugel_ is now known as kugel (
20:42:41 Quit kugel (Remote closed the connection)
20:42:50 Join kugel [0] (n=kugel@unaffiliated/kugel)
20:43:21 Quit FOAD (Read error: 110 (Connection timed out))
20:43:21 Nick FOAD_ is now known as FOAD (
20:46:17 Nick kugel is now known as kugel_ (n=kugel@unaffiliated/kugel)
20:46:18 Nick kugel_ is now known as kugel__ (n=kugel@unaffiliated/kugel)
20:46:39 Nick kugel__ is now known as kugel (n=kugel@unaffiliated/kugel)
20:50:33 Quit Horscht ("Verlassend")
20:50:33Unhelpfulit looks like the mac instruction has a "scale factor" for specifying a shift before adding... doesn't seem clear how that works :/
20:50:36 Quit MethoS (Remote closed the connection)
20:50:48rasherUnhelpful: preglow is our emac-ninja
20:50:56 Part yhuang
20:51:22Bagderstrong his powers are
20:51:40Unhelpfulrasher: that's the coldfire version, isn't it? i'm looking at the sh-1 docs right now. trying to figure out how to do 32x32->upper32 multiplies efficiently :/
20:51:52LearI did some emac-juggling too, but it was a while ago.
20:51:59rasherIgnore me
20:52:41Unhelpfulrasher: no, no, you're still actually being helpful, since it seems i'll need the same for coldfire as well
20:52:59Unhelpfularm is the only arch that hands us that operation in one instruction
20:55:47 Quit Thundercloud (Remote closed the connection)
20:55:55 Join casainho [0] (n=chatzill@
20:56:32 Join Horscht [0] (n=Horscht@xbmc/user/horscht)
20:57:08rasherI've been collecting a dataset of binary and ram usage sizes over time:"> - The data is nowhere near complete, and for this, build servers are need. Details are at the bottom of the page, but basically you need a linux server with git-svn and netcat. Network traffic is quite low, CPU usage is high
20:58:59Unhelpfulare those missing builds, or vertical lines that have been aliased out of existence?
20:59:07saratogaLear: replying to your message now
20:59:12jhMikeSUnhelpful: There are the FRACMUL macros in dsp.h but those don't cover SH.
20:59:15Unhelpful...i guess missing builds, seeing the horizontal ones.
20:59:16rasherUnhelpful: actually they're failed builds
20:59:30 Join tessarakt [0] (
21:00:28Unhelpfulhrm, is mac.l explicitly signed? :/
21:00:45jhMikeSIt depends on the emac mode.
21:01:19Unhelpful...ahh. so the unsigned version is the same, but i set a mode first? :/
21:01:41saratogaUnhelpful: the codec should all have a 32x32-64 bit coldfire macro
21:01:49saratoga"all the codecs"
21:02:38jhMikeSuse coldfire_set_macsr to set any emac modes. it returns the previous value. restore it when done.
21:03:31jhMikeSexcuse me,coldfire_get_macsr and coldfire_set_macsr
21:05:29 Join timc`` [0] (n=aoeu@
21:06:13jhMikeSI think unsigned multiply-accumulates might be int mode only, not fractional mode (it's been awhile).
21:08:48 Join jhulst_ [0] (n=jhulst@unaffiliated/jhulst)
21:09:51 Quit tim__ (Connection timed out)
21:10:00 Quit jhulst (Read error: 104 (Connection reset by peer))
21:11:05ajbCan anyone look at FS #9677?
21:11:14 Quit nibbler (Read error: 60 (Operation timed out))
21:15:36 Join lostnihilist [0] (
21:16:29 Join philips [0] (n=5563c62d@gateway/web/cgi-irc/
21:17:45lostnihilistI have auto-change directory set to yes; shuffle off; repeat off, only using the file browser, but when the current directory finishes, it doesn't move on to the next directory, it just stops playback
21:18:09lostnihilistiHP-120; suggestions on what I may be doing wrong?
21:18:17 Join Jaykay [0] (
21:18:20LloreanHow are you starting playback, and what SVN revision of Rockbox are you using?
21:18:38lostnihilisti browse through the "Files"
21:18:46 Quit shyam_k (Read error: 110 (Connection timed out))
21:18:49lostnihilistthen hold down Navi and "insert" into playlist
21:18:54LloreanDon't insert.
21:19:00LloreanJust click on the song you wish to start playing
21:19:14LloreanOnce you "insert" you're creating a playlist, rather than playing a folder.
21:19:21lostnihilistthat makes sense
21:19:23casainhohello :-)
21:20:00philipshi all
21:20:12casainhocan someone explain me why the needs of fliptable[] on firmware/drivers/ata-mmc.c ?
21:20:21philipsi have a philips hdd 1620
21:20:57philipscan anyone explain me rockbox is works with it or not?
21:21:39Bagderphilips: rockbox only works on the explicit devices mentioned on the front page
21:21:56Bagderthere are a few others in progress
21:22:33rasherIsn't the hdd1620 one of those in progress?
21:22:35philipsin introductions it seems everything works except "usb"
21:22:52Bagderwhat introduction?
21:23:04philipshow can you add sound and test it if usb not works?
21:23:18Bagderphilips: because the original OF still works and that has usb!
21:23:42philipssorry it is target status section
21:23:45JaykayjhMikeS: as i commented in fs 9812, everything works fine now. is it wanted that rockbox can shut down even if a usb-charger is plugged in?
21:23:57 Join Thundercloud [0] (
21:24:18Bagderphilips: ah right...
21:25:22Learcasainho: The MP3 decoder chip in the targets using that driver needs it.
21:25:31lostnihilistoh, ok, and you get playback on startup by setting the start screen to (under system) to 'resume playback'
21:25:34lostnihilistthat is rather obscure
21:25:47saratogaLear: you make it sound like fixing the AAC memory usage problems should be easy
21:25:59casainhoLear: can you please explain better?
21:26:01philipsso i can install it, can put music with its original OS, then listen with rockbox?
21:26:16 Part lostnihilist
21:26:24jhMikeSJayKay: You mean if you deliberately do so? I did enable that.
21:26:25 Join Comokanu [0] (n=5c0ce2b3@gateway/web/cgi-irc/
21:26:43casainhoLear: looks to me that responses from MMC card are a kind o fliped? no?
21:26:57ComokanuAre there any Rockbox simulators for MAc OSX?
21:27:08jhMikeSc200/e200 cannot tell the power sources apart at all anyway.
21:27:08philipsbadger: so i can install it, can put music with its original OS, then listen with rockbox?
21:27:37Jaykayi didnt get that, sorry
21:27:57 Quit philips ("CGI:IRC (EOF)")
21:28:11jhMikeSJayKay: You mean it's shutting down on its own or by holding the power button?
21:28:25Learsaratoga: Building the seek table isn't quite that easy, as you need to build it from 3 or 4 different tables, none of which you can load into memory completely.
21:28:27 Quit Comokanu (Client Quit)
21:28:38 Join philips [0] (n=55634196@gateway/web/cgi-irc/
21:28:53 Join nibbler [0] (
21:28:55Jaykayyes thats clear. i didnt get that with "tell the power source apart"
21:29:01saratogaLear: yes but samples table is the only one thats huge
21:29:05saratogathe others are quite small
21:29:47philipsafter installing rockbox , can put music with its original OS, then can listen with rockbox?
21:29:57philipsin hdd 1620
21:30:14jhMikeSJaykay: The same pin is used for USB/Charger power. There's no way to tell in software whether it's a USB charger or an A/C adapter.
21:30:48gevaertsphilips: possibly. I'm not sure if anyone currently here knows the details
21:31:03casainhosaratoga, Lear : are you talking about that fliptable I was asking about?
21:31:24Learcasainho: Hm, right, I was a bit too quick there, it seems. Amiconn is the one to ask, really.
21:31:40 Join itcheg [0] (i=41d59de2@gateway/web/ajax/
21:32:12philipsgevaerts: is rockbox utility installs for target-status players?
21:32:12casainhoLear: thanks :-)
21:32:25casainhoamiconn: are you there?
21:32:28 Join MethoS [0] (
21:32:32gevaertsphilips: rockbox utility only installs supported players
21:33:00 Join massiveH [0] (
21:33:00philipsgevaerts: so how can i try it?
21:33:12 Quit massiveH (Read error: 104 (Connection reset by peer))
21:33:15gevaertsphilips: as I said, I'm not sure if anyone currently here knows the details
21:33:56Jaykayjhmikes: im really sorry, i dont know what a a/c adapter is. let it be and stop confusing me :) i just wanted to know whether this behaviour (shutting down when a charger is plugged) is wanted
21:34:04saratogaLear: I just decoded a file without the samples table, now to figure out how to seek in one
21:34:16Jaykaybecause before your chnage it said nothing or "battery charging" and stayed on.
21:34:35philipsgevaerts: hmm, OK
21:34:46Learsaratoga: You can't rely on the other tables being small though. I've seen some pretty large chunk-related tables (hundreds or maybe even thousands of entries).
21:35:38saratogaLear: the chunk table can never be more then 2x the size of the samples table, and typically its much smaller so I'm not sure it matters
21:35:40jhMikeSJayKay: Allowing the user to power off even if charging was added with the charging commit for c200/e200.
21:35:44Learsaratoga: I'm pretty sure I have some test files that would break on it. :)
21:35:46saratogaa 33% savings is still huge
21:36:06*jhMikeS keeps capitalizing the 'K' for some reason...ah...hmmm.
21:36:41Jaykaythere are some more people who do that.... all of them are rockbox developers.
21:36:51Jaykaystrange isnt it?
21:37:22Jaykayis this usb-icon and the "battery: charging"-message still needed?
21:38:01 Quit philips ("CGI:IRC (EOF)")
21:38:18 Join BHSPitLappy [0] (n=BHSPitLa@unaffiliated/bhspitmonkey)
21:38:27jhMikeSthe usb icon is if connecting mass storage. Others still use the string.
21:38:30Learsaratoga: Which chunk table? There are two you know...
21:39:21Jaykayhow can a mass storage be connected to a mp3-player?
21:39:42 Quit jhulst_ (Read error: 104 (Connection reset by peer))
21:39:49 Join jhulst_ [0] (n=jhulst@
21:39:51 Join jimmy [0] (n=55634196@gateway/web/cgi-irc/
21:39:58jhMikeSAnother term for USB mode (MSC).
21:41:16Jaykayok now im totally confused. let it be, i wont understand it anyway:)
21:41:27jhMikeSok, np
21:41:35 Quit n17ikh (Read error: 104 (Connection reset by peer))
21:42:15 Quit jimmy (Client Quit)
21:44:01saratogaLear: you can't have more chunks then samples, so that puts a hard upper limit on the chunks table
21:44:41saratogaand the second chunks table [samples to chunks i think] always seems to be small, even for enormous mp4 video files
21:44:54 Join akur [0] (
21:45:00 Part akur
21:45:25 Join gregzx [0] (
21:45:37Learsaratoga: True, but that still means thousands of entries. Especially for audio books, with several hours per file.
21:46:28Learsaratoga: Also, to reduce the sample table (or convert it to seek points), you need the chunk offset table, so you don't merge a gap between frames.
21:46:37 Join tim__ [0] (n=aoeu@
21:47:33 Quit BigBambi (Read error: 60 (Operation timed out))
21:47:34saratogathousands of entries isn't a big deal
21:47:49saratogathe current code seems to use thousands per minute
21:48:11 Join BigBambi [0] (n=Alex@rockbox/staff/BigBambi)
21:48:16saratogathousands per hour sounds like a great improvement :)
21:49:27saratogaspeaking of which, does anyone here actually have vorbis files using floor0? I'd love to see how much memory those use verses modern vorbis
21:49:43Unhelpfuldo i need to reset the emac mode before yielding? or would setting it once at the beginning of the scaler, and setting it back to its previous state at the end, be fine?
21:50:20Unhelpfuland i'm assuming "integer mode" means than mac.l is an unsigned 32x32 multiply?
21:50:27***Saving seen data "./dancer.seen"
21:50:47Learsaratoga: I have a one or two floor0.
21:51:00saratogaLear: could you post them somewhere?
21:51:46Learsaratoga: It is less than 1 MB, so maybe I could mail it?
21:51:47jhMikeSUnhelpful: No, macsr is saved in thread context
21:52:47Unhelpfulok... so basically, i only need to reset it before returning, so that it's in whatever mode the calling thread expects
21:53:07 Quit timc`` (Connection timed out)
21:53:44 Join yhuang [0] (
21:54:01saratogaLear: did you get my message?
21:54:21jhMikeSUnhelpful: just restore it once you've left a call. it gets a DSP-compatbile setting by default.
21:54:52Learsaratoga: Got it now.
21:55:51saratogaI wonder how important floor0 support is, as far as I know it hasn't been used in vorbis encoders in over 6 years
21:56:09saratogathough maybe it doesn't use much memory
21:56:50mcuelenaerecasainho: I could be wrong here, but I think ata-mmc.c isn't a really generic driver; so you should make your own if you're trying to use it
21:57:05 Quit gregzx ("ChatZilla 0.9.84 [Firefox 3.0.5/2008120122]")
21:57:26 Join gregzx [0] (
21:58:24casainhomcuelenaere: yes... but I can't get my drives working and I don't know If the my problem is not being doing that "fliptable", so; I would like to know...
21:58:47casainhoI hope amiconn can help me...
21:59:35 Quit TheSphinX^ ("XChat@Linux")
22:01:20mcuelenaerecasainho: how exactly is your SD card slot connected to your CPU? (I can't seem to find that on the wiki page)
22:02:00 Join an|noid [0] (n=std~den@
22:02:01 Join petur [50] (n=petur@rockbox/developer/petur)
22:02:31casainhomcuelenaere: here:
22:03:36casainhomcuelenaere: it's is connected by SPI and by SD bus, however there I would prefer to use SPI since looks like there is not royalties to be paid if we use SPI... but we need to pay to use SD bus :-(
22:04:01mcuelenaereare there any drivers provide by atmel for SD over SPI?
22:04:13casainhomcuelenaere: no....
22:04:44casainhothere is a lot of examples for SPI, for others MCUs as AVR and PICs...
22:04:52casainhoand also ARM7, for example
22:05:15casainhohere: file:///home/cas/Documentos/rockbox_player/03_info/sd_card/electronic-atmega-mmc.php.html
22:06:13mcuelenaereI'm not going to be able to get a document from your PC without you uploading it first ;)
22:06:47 Join n17ikh [0] (n=n17ikh@
22:06:49casainhomcuelenaere: sorry:
22:07:44 Quit J-23 ("ZNC -")
22:07:59 Join J-23 [0] (n=zelazko@
22:08:26mcuelenaerecasainho: do you already have a working SPI driver?
22:08:52casainhomcuelenaere: yes, I have, I tested it with osciloscope... and even I have it working for the LCD ;-)
22:09:54casainhoand I am getting answer from the card, however seems witout sense :-( −− that's why I am looking at that fliptable[]...
22:10:40casainhoyou know, I am doing JTAG debug and I can also measure the signlas with osciloscope ;-)
22:11:05 Quit an|noid (Read error: 104 (Connection reset by peer))
22:12:02casainhomcuelenaere: for example, after "sdcard_send_command (CMD_GO_IDLE_STATE, 0x0, 0x95);" I am always getting the response "0x11111110" :-(
22:12:22casainhomcuelenaere: while it should be 0x01
22:13:23mcuelenaerehmm I can't really help you with that, as I'm not familiar with SD (haven't done a (working) SD driver yet)
22:14:20casainhomcuelenaere: but, did you try to make any SD card driver?
22:15:12mcuelenaereyes, but it was more a copy-and-paste job and I haven't really looked at it (yet)
22:15:48casainhookok :-)
22:18:22 Quit Lear ("ChatZilla 0.9.84 [Firefox 3.1b3pre/20090116035118]")
22:19:48 Join faemir [0] (
22:21:18 Join japc_ [0] (
22:27:52 Join gregorovius [0] (
22:29:53 Join Jaykay_ [0] (
22:30:30 Quit XavierGr (Nick collision from services.)
22:30:41 Join XavierGr [0] (n=xavier@rockbox/staff/XavierGr)
22:30:54 Quit casainho ("ChatZilla 0.9.84 [Firefox 3.0.5/2008121622]")
22:31:01 Join MethoS- [0] (
22:32:47 Quit karashata ("G'bye everyone!")
22:36:41 Join Seed [0] (
22:37:00 Quit {phoenix} (Read error: 104 (Connection reset by peer))
22:37:07 Join n17ikh| [0] (n=n17ikh@
22:37:14 Quit n17ikh| (Remote closed the connection)
22:40:07 Join MethoS-- [0] (
22:40:37 Join casainho [0] (n=chatzill@
22:41:35 Nick jhulst_ is now known as jhulst (n=jhulst@unaffiliated/jhulst)
22:44:48 Quit n1s ()
22:45:32 Part Anges
22:46:07casainhoamiconn: are you there?
22:46:15 Quit MethoS (Read error: 110 (Connection timed out))
22:46:58 Quit n17ikh (Connection timed out)
22:48:14Unhelpfulgevaerts: would you mind testing some coldfire-specific scaler stuff on your x5?
22:48:44 Quit Jaykay (Read error: 110 (Connection timed out))
22:49:20gevaertsUnhelpful: I can do x5 or h300
22:49:53Unhelpfuli've a build for x5 already, but i can make one for the h300 as well
22:51:58gevaertsYour choice
22:53:20Unhelpfulanything that uses scaled bitmaps should be fine for a test... pictureflow, albumart if you don't have it pre-scaled to a WPS size, sliding_puzzle is good, too
22:53:26amiconnUnhelpful: macl/mach cannot be shifted
22:53:29 Quit japc_ (Read error: 110 (Connection timed out))
22:53:55amiconnThe sh programming manual is also in the wiki
22:54:31Unhelpfulamiconn: i had figured that might be the case. and i'm guessing the "scale factor" for the mac instruction probably doesn't offer convenient values?
22:55:21 Quit Jaykay_ ("ChatZilla 0.9.84 [Firefox 3.0.5/2008120122]")
22:55:50 Quit MethoS- (Connection timed out)
22:56:46 Quit jhulst (Read error: 60 (Operation timed out))
22:56:54 Quit archivator ("<>")
22:57:55 Join Zagor [242] (
22:58:42gevaertsUnhelpful: lots of coloured noise
22:58:51Unhelpfulfantastic. :/
22:59:01 Quit XavierGr ()
22:59:02 Quit petur ("Zzzzz")
23:01:51casainhoamiconn: can you help me on SD SPI drivers?
23:03:19 Quit bmbl ("Woah!")
23:04:18 Quit tessarakt (Read error: 110 (Connection timed out))
23:04:49casainhoamiconn: I would like to know why the fliptable[] on drivers/ata-mmc.c ...
23:04:52Unhelpfulwell, it seems that at *least* i have the emac mode wrong
23:06:07Bagdercasainho: that's archos-specific fiddling
23:06:59casainhoBagder: fiddling?
23:07:13*Unhelpful sees that the "scale factor" part was from the coldfire version. meh.
23:07:29casainhoBagder: I don't understand why an SPI connection would change the bits.... ?
23:07:45Bagdercasainho: because that's needed on the archos
23:08:08Bagderit's not strictly related to spi
23:09:15casainhoBagder: okok...
23:10:28 Quit casainho ("ChatZilla 0.9.84 [Firefox 3.0.5/2008121622]")
23:11:09 Join arohtar [0] (
23:11:14 Join tessarakt [0] (
23:12:22 Quit amiconn (Read error: 110 (Connection timed out))
23:12:26 Quit pixelma (Read error: 110 (Connection timed out))
23:12:51 Join pixelma [0] (n=pixelma@rockbox/staff/pixelma)
23:13:01 Join amiconn [50] (n=jens@rockbox/developer/amiconn)
23:13:17 Join shfhs75 [0] (
23:14:02shfhs75iam man frome greece iny girl hir??????
23:14:18Zagorgo away
23:14:32amiconnUnhelpful: You're not obliged to make it perfect in the first commit
23:14:41shfhs75fuck you man
23:14:48Mode"#rockbox +o Zagor " by ChanServ (ChanServ@services.)
23:14:52Kick(#rockbox shfhs75 :Zagor) by Zagor!n=bjst@rockbox/developer/Zagor
23:15:01amiconnIf you don't have the necessary hardware to test an asm optimisation on, someone else should probably do that optimisation
23:15:26amiconnThe emac scale factor can only do >>1 or <<1, and only in integer mode
23:15:42Mode"#rockbox -o Zagor " by Zagor (n=bjst@rockbox/developer/Zagor)
23:15:53Unhelpfulisn't integer mode what i want, for unsigned?
23:16:21amiconnIf we want to "misuse" the emac for an unsigned 32*32->high 32, we need some clever tricks.
23:16:52amiconnemac can be used signed or unsigned in integer mode, but only signed in fractional mode (which you need to get to the high bits)
23:17:10Unhelpfulahhh.. that's problematic. but maybe not. :)
23:18:13Unhelpfulthe scale factor only has 31 bits of precision, to avoid needing a 64-bit division to calculate it
23:19:13amiconnWell, if we can shift both factors right by 1, it becomes fairly simple
23:20:57Unhelpfulwe can already shift ctx->divisor right by one, it's calculated as (0x80000000/div) << 1 to avoid a 64-bit divide
23:21:25 Nick JdGordon|zzz is now known as JdGordon (n=jonno@rockbox/developer/JdGordon)
23:22:26Unhelpfulthe other factor, we may not need to... it is, at maximum, equal to 255*w*h, where w and h are the larger of the source or target dimension in that direction.
23:23:14 Join MethoS [0] (
23:24:16amiconnHmm, if the other factor is guaranteed to stay <2^31, it'd be even better.
23:24:17Unhelpfulso, it will be 31 bits or less all the way up to 2901x2901
23:25:19amiconnemac in fractional mode actually gives you bit 31..62 in the accumulator, so if one factor is shifted right by 1 and the other is guaranteed to stay positive, the result won't need shifting at all
23:25:23Unhelpfulwell, we can remove the left shift when calculating the divisor, and the other factor is guaranteed to be <2^31 if the source bitmap has less than about 8.4 million pixels
23:25:44amiconnThe question is whether we can drop the calculation of the low bits. We'd loose a bit of precision
23:26:10Unhelpfulnot sure what you mean by "drop the calculation of the low bits"?
23:27:04amiconnFor a true 32*32->64, we'd need a mulu.l (producing the low bits)and a mac.l (producing the high bits, with emac in fractional mode)
23:27:40Unhelpfulthe low bits are discarded entirely, though...
23:27:41amiconnSince we only need the upper 32, all the low bits would contribute would be a bit of rounding
23:28:04amiconnIf we drop them, we don't even need to perform the mulu.l
23:28:56Unhelpfulthey're already dropped entirely on arm. unless you're saying that the top 32 will lose some precision if we don't have a mulu.l to produce the bottom 32?
23:29:29amiconnBtw, there's a trick for reconstructing bit 63 without accessing the accumulator extension register. We won't need that here though
23:30:02amiconnNo, just that you always round towards zero if you don't calculate the bottom 32 bit
23:30:21 Join toffe82_ [0] (
23:30:38Unhelpfulwell, we actually add a rounding value before the multiply, at least on arm
23:31:10Unhelpfuland just discard the bottom 32 entirely
23:31:22 Quit domonoky (Read error: 104 (Connection reset by peer))
23:33:39Unhelpfulctx->round is half of the divisor (from before the reciprocal is calculated), so the (val + round) * recip >> 32 is (val + div/2) / div
23:35:11 Quit nibbler (Read error: 110 (Connection timed out))
23:36:24Unhelpfulctx->divisor is only truly the divisor until the top of scale_v_*, at which point it's replaced with (2^31 / divisor) << 1
23:37:33Unhelpfulso, it sounds as if, if we remove the <<1, mac.l and extracting the high bits gets us exactly the same result we have on arm... provided the image is never large enough for the pixel value to exceed 2^30
23:38:08 Quit MethoS-- (Connection timed out)
23:44:22 Quit MethoS (Remote closed the connection)
23:44:46 Join MethoS [0] (
23:45:07 Quit Aurix_Lexico (Remote closed the connection)
23:47:13 Join Aurix_Lexico [0] (
23:47:50 Quit toffe82 (Read error: 110 (Connection timed out))
23:48:16 Join MethoS- [0] (
23:50:30***Saving seen data "./dancer.seen"
23:50:32Zagoramiconn: how did you measure the coldfire PLL locking time? it just struck me maybe we get equally fast clock switches with PLL bypass if we just don't change the VCO.
23:52:39Unhelpfulamiconn: if i'm reading the programmer's reference correctly, for the normal scaling case, i can actually use unsigned integer mode, with the desired result being in the lower extension byte
23:53:21 Quit tessarakt ("Client exiting")
23:53:38Unhelpfulfor the parcticular case that is being used in pictureflow, of rescaling the existing scale factors by 31/255 and 63/255, it's a little more complicated, because i want a 32-bit value back
23:53:42amiconnZagor: Hmm, did I say I measured it? I can't remember atm. Didn't Linus do that?
23:54:12Zagoroh. I'm not sure.
23:54:43Unhelpfulperhaps the right thing to do is to give a custom output plugin a function that can be called once, at the start of the image, to calculate its own scaling factors, or whatever else it might need.
23:54:54amiconnI think it could be measured using the user timer, but you need to accept a bit of uncertainty
23:55:11ZagorI'll see if a simple loop counter shows any difference
23:55:26 Join jhulst [0] (
23:55:31amiconnYeah, you can count loop iterations while waiting for the relock bit
23:55:55 Join gregorovius_ [0] (
23:56:32amiconnThe problem is how to translate this into cpu cycles. There are lots of waits involved in mbar/mbar2 accesses
23:56:53 Quit arohtar (Client Quit)
23:56:59amiconnHmm, it can be calibrated separately
23:57:07 Join arohtar [0] (
23:57:53Unhelpfulif i'm using %acc0, it looks like the 8 bits i want would be the low 8 of %accext01
23:58:32amiconnYou still have to make sure acc0 is cleared afterwards

Previous day | Next day