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 2009-12-05

00:08:56 Part captainkwel
00:17:14 Join Tomis2 [0] (n=Tomis@
00:17:58 Join stripwax [0] (
00:18:22 Quit Tomis (Read error: 60 (Operation timed out))
00:18:23 Nick Tomis2 is now known as Tomis (n=Tomis@
00:19:35stripwaxmt - ping
00:24:50n1syay, division by 0
00:27:20 Quit faemir ("Leaving")
00:28:14 Join FOAD_ [0] (
00:29:49LambdaCalculus37pixelma: ping
00:31:27 Join NoGare [0] (i=d80887b8@gateway/web/freenode/x-xidfeudrdwzrqauy)
00:32:41 Join drf|laptop [0] (
00:33:27drf|laptopHey guys. So I have a Sansa e200, and it has having some oddities with some Rockbox files being corrupted and freezing explorer any time I tried to copy/move/erase them. So I did a complete uninstall of Rockbox, and reinstalled it. I copied the themes and stuff back but now I'm seeing something weird.
00:33:44drf|laptopwhen I go to the file browser, it's like... some stuff's yellow, some stuff's green, some stuff's red, etc.
00:33:59drf|laptopit was never like that before - just solid black for everything. Any idea why it's doing that? I can't even find a setting to change it
00:34:17stripwaxmaybe it's the theme you've installed?
00:34:34drf|laptopI don't think so... I did "Reset colors" and it still does that
00:34:49stripwaxif you select the Cabbie theme, does the file list look like you expect?
00:34:53drf|laptophold on
00:35:01drf|laptopI just switched to some other random thjeme and it still does that
00:35:18stripwaxok but what about Cabbie theme ..
00:35:29drf|laptopNope. Still different colors
00:35:31drf|laptopon Cabbie
00:35:45gevaertsyou loaded a theme that sets custom file colours, and those are sticky
00:35:53drf|laptopI can take a picture if you want... but it just seems like each filetype has its own color
00:36:05drf|laptopis there a way to unload it?
00:36:17stripwaxdrf|laptop : - what gevaerts said (although, I thought changing theme reset them .. )
00:36:28drf|laptopI changed to Cabbie and it's still custom
00:36:41gevaertsstripwax: it doesn't. There was talk about that in here a few days ago
00:37:13drf|laptopliterally, all I did was this. Reinstall Rockbox, copy back the .rockbox files from a backup, saying "no" to overwriting anything there (basically, putting back my added themes and stuff)
00:37:21drf|laptopthen just overwrote the .conf with my settings
00:37:27drf|laptopand it started all funky when it never did that before
00:37:47stripwaxgevaerts - yep (so I think I misremember the outcome). Is that going to be fixed or was it decided that the current behaviour is desirable?
00:39:34 Quit dutchie ("goodnight")
00:41:49 Quit FOAD (Read error: 110 (Connection timed out))
00:41:49 Nick FOAD_ is now known as FOAD (
00:45:20 Quit dmb ("Leaving")
00:50:43drf|laptopanywa... any idea how to fix the custom colors?
00:57:26 Quit p3tur ("*plop*")
00:59:50 Quit petur ("Zzzz")
01:07:20 Part toffe82
01:15:15 Part drf|laptop
01:21:00 Quit stripwax ("")
01:23:24 Quit evilnick ("Page closed")
01:28:40 Quit geertvdijk (Read error: 110 (Connection timed out))
01:33:54n1syay i reproduced the crash on sim
01:41:36Unhelpfulcrashes you can't reproduce on sim make me wish for gdb on target :)
01:45:57n1sthis one is pretty straight forward actually, clamp value to lowest possible fixed point number larger than 0, multiplu that with a fixed point number slightly less than 1, divide by the result, bang
01:46:11n1sonce you see past the fixed point black magic
01:46:17 Join faemir [0] (n=faemir@
01:47:32n1sso clamping to a slightly larger value fixes it :)
01:48:45 Join DerPapst [0] (
01:52:14n1sthis pitch detector plugin could probably be sped up greatly on coldfire with some iram usafe
01:52:23n1susage, even
01:59:03 Join CaptainKewl [0] (
02:00:02***Saving seen data "./dancer.seen"
02:00:49CIA-6New commit by nls (r23850): Fix division by 0 by clamping freq value a little higher, some minor style and whitespace changes
02:03:45Unhelpfuland its use is presumably already exclusive with playback, anyway?
02:04:40n1sUnhelpful: i think so, it records from mic/line in...
02:05:15Unhelpfulstrict aliasing seems to have the same effect on e200 as on beast, of making -O3 best *much* more often
02:05:34n1snice :)
02:06:26Torneyah, aliasing often prevents inlining in cases where it is otherwise reasonable
02:06:36Unhelpfulit looks as if, on ARM, we can really just set -O3 for codecs, with a few exceptions that can use -O2
02:06:54Tornegcc makes surprisingly good inlining decisions these days in general
02:08:17Unhelpfuli think it might be worth using a clz based on our core asm ffs implementation on coldfire - this can be done branchless, although the "reduce-to-first-set-bit" logic is quite a bit more work on an arch without shifted operands
02:08:28 Quit Casainho (Remote closed the connection)
02:08:34 Join kugel_ [0] (
02:09:57n1sUnhelpful: nice, i wanted to take a look at how libfaad is calculating log2, saratoga mentioned it was a heavy user and a speedup there would be much more worthwhile than flac :)
02:10:54n1si also wonder if a special case for 0 should be added fo the trivial clz log2 i hacked into flac as that will return -1 for 0 while the original av_log2 returns 0
02:11:18n1sthe output was unchanged for my testfile but i don't know if this is unsafe
02:13:18Unhelpfuli must confess that i have *no idea* how floor_log2 works
02:14:26Unhelpfulwait, yes i do... it only *looks* like the reduce-to-first-bit logic, it fills everything below the highest set bit with 1s and then does a population count.
02:14:50 Join johnnynm [0] (n=Administ@
02:15:42Unhelpfulthe popcount logic uses lots of 32-bit literals that can't be immediate on ARM. i would expect this to be a terribly pessimistic implementation, although it can probably be done efficiently as thumb code
02:18:35 Quit ps-auxw (Read error: 110 (Connection timed out))
02:19:41n1suh, i wonder if this pitch detector thingy should stop playback before starting to mess with the pcm settings and whatnot
02:20:05Unhelpfulplugins that use iram will stop playback before starting anyway, won't they?
02:21:02n1syes, but this one doesn't use iram (yet)
02:22:08n1sthere are crashes reported in the tracker when stating the plugin while playback is running, works fine on my h300 but may be target specific of course
02:23:38n1si should find my sansa cable
02:33:52Unhelpfulah, you didn't commit its use of iram ;)
02:35:11Unhelpfulfaad also doesn't use the clz instruction at all ;)
02:38:58 Join TopyMobile [0] (
02:41:49 Quit DerPapst ("Leaving.")
02:55:04n1sspeaking of faad, we are still based off of that pre gpl incompatible license change version, but they since changed the license back to gplv2 so maybe we should look into merging in and interesting upstream changes too?
02:55:53 Join fdinel [0] (
02:58:42saratogan1s: faad uses log2 a lot but i don't know if its performance critical or just maybe to compute lookup tables
03:00:21n1ssaratoga: they seem to use both a function and tables, no idea if they are performance critical or not
03:00:31n1si guess a benchmark will tell me :)
03:01:42n1ssaratoga: is sbr only used in HE ?
03:01:43 Quit johnnynm (Read error: 110 (Connection timed out))
03:02:40saratogan1s: yes SBR is just HE
03:02:47saratogaand IMO is most in need of optimization
03:04:11n1sonly files that are named sbr* seem to use the functions that can benefit from a faster clz so i guess that if it will help speed it will be for HE
03:08:00 Quit hebz0rl ("Ex-Chat")
03:10:56Unhelpfulif aac is abusing log2 heavily then we really could use a better, and inline, log2
03:11:26n1sit doesn't *look* heavy to me but only a test will tell
03:12:48Unhelpfulit looks like floor_log2 returns 0 for x=0. on armv6 this is easily done as __builtin_clz(x) & 31
03:13:47n1sUnhelpful: ah, didn't think of that
03:14:09n1snot needed here though as the only calling places are already guarded
03:15:44n1sbut i will use that in flac then to be on the safe side, shouldn't be much more expensiv
03:16:42Unhelpfuln1s: in that case i'd suggest we use inline __builtin_clz for floor_log2, with an alternate function for architectures where that makes sense
03:17:55Unhelpfulthis one is 10 instructions, one of which is a load, none conditional... and shouldn't stall arm7 or arm9 pipelines at all:
03:18:51 Quit LambdaCalculus37 ("")
03:19:59Unhelpfulit can't be inlined in that form due to the hack used to address the LUT. i'm not sure it would make sense to inline it? inline, it would need another load to fetch the table address... probably costs about the same as a branch, but perhaps not when you account for register saving.
03:22:10n1sbut it uses only one reg, not much to save
03:23:09n1salso i don't really follow you algo, i get the set all lower bits and isolate the msb but not the rest
03:23:18Unhelpfuln1s: gcc would save all caller-saved regs, wouldn't it?
03:23:48n1syou probably know more about that than i fo :)
03:25:27n1swhen i think about it, yes it has to save those but aren't there like just 2 on arm?
03:25:59Unhelpfulthe rest is taken from ffs-arm.S. the shift sequence is a multiply in disguise - this is equivalent to shifting the multiplier, since there is only one bit set. the trick is that the top 6 bits of the multiplier are unique for each shift position.
03:26:24Unhelpfuln1s: r0-r3, ip, and lr, i believe.
03:26:38 Quit NoGare (Ping timeout: 180 seconds)
03:27:35Unhelpfuland the last instruction before return branch is a trick. you can use PC as the base register for a LUT *if* the LUT is at current instruction + 8
03:28:41n1sheh, pretty sneaky
03:30:00 Join ps-auxw [0] (n=arneb@2001:470:c807:0:1532:4e5f:2ad3:4123)
03:31:04Unhelpfulwe could reduce register saves to 3 if we cheat a bit - make an inline asm macro for floor_log2 that calls the actual function, and explicitly clobbers lr. it would basically have an input/output value in r0, clobber lr, and consist entirely of "bx floor_log2_impl" :D
03:33:16 Quit n1s ("Lämnar")
03:34:00Unhelpfulhrm, that table is also not "right" for floor_log2, of course.
03:34:28 Join kugel [0] (n=kugel@rockbox/developer/kugel)
03:36:47 Quit grndslm (Remote closed the connection)
03:39:02 Quit efyx_ (Remote closed the connection)
03:49:34Mode"#rockbox +o Unhelpful " by ChanServ (ChanServ@services.)
03:49:45Mode"#rockbox -o Unhelpful " by ChanServ (ChanServ@services.)
03:51:21 Quit kugel_ (Read error: 110 (Connection timed out))
03:52:12 Join toffe82 [0] (
04:00:04***Saving seen data "./dancer.seen"
04:14:16Byandid I screw something up, or is this "multimedia mode" new in bleeding edge?
04:16:30saratogaits been in for a while
04:16:44saratogaeven the old 3.4 build has it
04:17:04Byansaratoga: how do you get to it in the 3.4 build?
04:18:05Byancause I just had the 3.4 build on here, and it was definately using UMS
04:18:32UnhelpfulByan: it still does, it just *also* exposes the buttons to the attached host PC as an HID device
04:18:40 Join fyrestorm [0] (
04:19:16Unhelpfuland i think that perhaps multimedia mode might have been the only HID mode as of 3.4, and not used a label?
04:19:55Byanerr.. ok
04:20:36*Byan made it so that it charges by default, but when he tested it, it didn't boot UMS, so he was confused at first
04:21:37Unhelpful"multimedia mode" is still UMS also. and should be charging at the same time, if your device can charge over USB in rockbox (some don't do that very well)
04:22:26Byancharge only by default?
04:22:40Byanthats what I mean
04:22:50Byansorry, let me reiterate
04:23:04Byanwhen I plug my ipod in, I usually want to continue to listen to muisc
04:23:10Byanand charge it
04:23:20Byanso now, when I plug it in, it charges and continues playing..
04:23:32ByanI have to hold menu in order to start multimedia mode
04:24:13saratogayou compiled your own build?
04:24:13Unhelpfulok, so it's charge-only, by default. and you have to hold menu to use UMS. i'm not quite sure what the question is?
04:25:17Byansaratoga: yeah
04:25:20Byanthere isn't any question
04:25:30ByanI was just really confused because it did the multimedia mode
04:25:42ByanI was using 3.4 before this
04:25:42Unhelpful"multimedia mode" is still UMS. :)
04:26:05Byanwell, what's it called when it reboots into the apple looking transfer thing?
04:26:15Unhelpfulin 3.4 support was added for controlling the PC while in UMS mode, using the rockbox buttons.
04:26:50saratogathat would be booting into the Apple firmware
04:27:03Unhelpful*since* 3.4 support was added for having more than one control scheme for the device buttons while connected, and text is displayed below the plug icon indicating which mode the buttons are in
04:27:31Unhelpful"multimedia mode" means that the buttons on your ipod will look like media-control keys (stop, pause, mute, volume up/down, etc)
04:27:50Byansaratoga: alright, so, why did it stop doing that..
04:28:17ByanI'm obviously misusing terms
04:28:27Byanbefore when I plugged it in, it used to reboot into the apple looking transfer thing
04:28:27saratogawe have rockbox USB now
04:28:38saratogaso we don't need to reboot into the apple firmware to use USB
04:28:47Byanand thats new since 3.4?
04:29:03saratogaits been in svn forever
04:29:09saratogai don't remember which build first had it enabled
04:29:23Byanhrm, weird, cause it's not in 3.4 >_>
04:29:26Byanok, whatever
04:29:34Byaneverything works, so, I'm happy
04:29:46saratogait may have been disable din 3.4
04:29:55saratogaactually i'm not even sure if its going to be in 3.5
04:30:03Byanwhy is that?
04:30:15Byannot stable?
04:30:19saratogai don't know
04:35:32 Nick krazykit` is now known as krazykit (n=kkit@
04:37:39Byansaratoga: are there any other things that are disabled in releases?
04:40:04UnhelpfulByan: anything that's incomplete or unstable *might* be disabled in release builds
04:40:27Unhelpfulsaratoga: thanks for the catch, i forgot about rockbox USB being disabled in 3.4... that was only for certain PP targets, right?
04:44:23saratogai think it was just ipods since we don't have charging on some of them, but i didn't follow that closely
04:45:13*Byan often forgets rockbox runs on things other than ipods
04:45:20Byansince it just runs so well on them =p
04:48:30Unhelpfuln1s: for the logs: the thing using a table is a fixed-point 2^n function, i think. floor_log2 is used only by log2_int and log2_fix, and log2_fix is never even used, it appears. there's more than a little mess in here. :)
04:51:46 Join dmb [0] (n=Dmb@unaffiliated/dmb)
04:51:54Unhelpfulwl_min_lzc could be reduced to clz + some fixup instructions, too... and the table version could entirely remove the fixup. also the return value *always* has 1 subtracted if non-zero. for table-based versions this could be done by adjusting the LUT values, so moving the x = x ? x-1 : that always follows it into the function could be a benefit.
04:53:04FlynDicetopik: A couple more fuze reports coming in now, I don't think it's just your card.
04:54:27FlynDicetopik: Trying to gather more info now
04:59:21 Join webguest05 [0] (
05:00:08webguest05question about the gigabeat s & rockbox
05:00:39webguest05there anyway to get a smaller, only booting rockbox nk.bin from an already patched nk.bin?
05:00:41 Join angelwolf71885 [0] (
05:00:52 Quit webguest05 (Client Quit)
05:00:55 Join webguest54 [0] (
05:00:55webguest54there anyway to get a smaller, only booting rockbox nk.bin from an already patched nk.bin?
05:01:48webguest54well yea, anyway to get a nk.bin that only runs rockbox from a nk.bin that dual boots?
05:02:08webguest54(bought it preinstalled with rockbox, and i dont care to use the OF)
05:04:27Unhelpfulwebguest54: you could build one from the rockbox source, but it is not guaranteed to work. certain beast flash versions will force restore if you use a non-dual-boot rockbox.
05:04:43 Quit dmb (Read error: 104 (Connection reset by peer))
05:09:15webguest54what about the nk.bin thats on the server (trying to find the link again)
05:10:40webguest54the online html help things down/gone
05:10:48webguest54so i cant find it :(
05:10:58webguest54hopefully you guys know what im talking about
05:17:46Unhelpfulthere's no tool available to extract a single-boot bootloader from the dual-boot one.
05:19:40webguest54uhm, this might go against your "rules" but would it be possible to just upload a nk.bin thats single-boot?
05:20:17webguest54(i figure it might "infringe" on M$s "intellectual property" or something)
05:21:09Unhelpfulno, it wouldn't infringe. if anything, the dual-boot nk.bin infringes, as it contains the original nk.bin
05:23:01webguest54well then, would anyone mind uploading a single boot nk.bin? (or would that not work as maybe the NK.bin has to be patch with your players serial #?)
05:24:21Unhelpfulit doesn't need to be patched, but to upload one, someb ody would have to build one. also that person might be seen as responsible for any problems if it didn't work.
05:25:01webguest54i wouldnt care, i already backed it up anyway
05:25:04webguest54but whatever
05:27:23 Quit webguest54 ("CGI:IRC (EOF)")
05:32:51 Quit angelwolf71885 ("ChatZilla 0.9.85 [Firefox 3.5.5/20091102152451]")
05:37:55 Quit fdinel ("Miranda IM! Smaller, Faster, Easier.")
06:00:07***Saving seen data "./dancer.seen"
06:06:30 Quit reid02 (Read error: 113 (No route to host))
06:11:47 Join Tomis2 [0] (n=Tomis@
06:13:30 Join Exiles [0] (
06:18:25 Quit kugel (Read error: 113 (No route to host))
06:20:50 Quit Tomis (Read error: 110 (Connection timed out))
06:20:50 Nick Tomis2 is now known as Tomis (n=Tomis@
06:21:36 Join GeekShadow [0] (n=Antoine@reactos/tester/GeekShadow)
06:22:53 Quit GeekShado_ (Read error: 104 (Connection reset by peer))
06:46:50topikFlynDice: would it still be useful to try older builds?
06:47:48topiki'm building r23828 and r23807 now to test. both are from just before AMS changes
06:47:57FlynDicetopik: I've got an idea I'm working on right now
06:48:33 Quit mt (Read error: 104 (Connection reset by peer))
06:49:13FlynDiceI'm thinking perhaps the init is going wrong when we switch the cards to HS timings, you going to be around for a bit?
06:49:45 Join mt [0] (n=mtee@rockbox/developer/mt)
06:51:43FlynDiceI don't think you should waste time testing the older builds.. I'm pretty sure about the one that broke it but I don't think it'd breaking where the change was made.... can I have you try a patch in 5 mins or so?
06:52:09topikyeah, i can try anything you want :)
06:52:22FlynDicethanks, be with you in a few mins
06:53:50Exilesokay, I have a question. I have an ipod mini 1g, today I was given an inline RF remote control. it plugs into the jack on the top. what should I look into modifying if possible to add serial access via the top port so I can try to hack in accessory support for it? I can debug the rx/tx lines if that helps
07:01:01 Join angelwolf71885 [0] (
07:01:46Exilesforget a /j there angelwolf71885?
07:02:00angelwolf71885possably lolz
07:02:10Exilesno worries
07:02:57angelwolf71885so how is everyone tonight/day?
07:03:11ExilesI'm just trying to figure out how to enable serial on my ipod mini so I can hack an AAP device into it
07:03:27FlynDicetopik: Could you try this and see if it helps
07:03:30angelwolf71885humm nt fully sure
07:03:46angelwolf71885buti barely understand most of this stuff
07:03:53 Join bluebrother [0] (
07:04:12angelwolf71885im waiting for linux4nano to git exicushion on the nano 3g xD
07:05:01angelwolf71885what is an AAP device btw?
07:05:08 Quit bluebroth3r (Read error: 60 (Operation timed out))
07:05:11Exiles I'm working with that, I can easily debug the rx/tx lines
07:05:12angelwolf71885is that like dcking it to a sterio?
07:05:22Exilesapple accessoriy device
07:05:51angelwolf71885ok a dck device like i thought xD
07:06:01angelwolf71885sorry my O is busted
07:06:10Exilesjust via the 4 pin conector on top
07:06:23angelwolf71885never wned a mini xD
07:06:25Exilesvs the 30 pin ipod jack
07:06:36Exilesit was free minus the cost of a battery
07:06:54angelwolf71885your remote?
07:06:58angelwolf71885or the mini?
07:07:06Exilesthe ipod mini, and the remote
07:07:13angelwolf71885LOL score
07:07:28Exilesagreed tossed a 16gb CF card in it
07:07:33angelwolf71885did you mod the mini for a CD drive yet?
07:07:53Exiles4gb was to small
07:08:04angelwolf71885i know what you mean
07:08:06Exilestraded the 4gb hdd for a voip sip device
07:08:12angelwolf71885i own the low level nano 3g
07:08:25angelwolf71885but 4Gb is perfict for 24 hours of tunes
07:08:31Exilesthen I was given a busted ipod 5g 30gb device
07:08:40angelwolf71885o wow very nice
07:08:53Exilesonce i powered it up via firewire it booted up
07:08:57angelwolf71885my csion owns a 6.5 classic
07:09:04angelwolf71885the hd is dead
07:09:10Exilesthen I replaced it's dead battery and it's busted audio jack
07:09:21angelwolf71885we wanna add a CF sled to it and git it back up and running
07:09:31angelwolf71885very nice
07:09:39angelwolf71885your one luckey duck
07:09:40 Quit Tomis (Read error: 110 (Connection timed out))
07:10:18Exilesthey were all broken when I got them
07:10:29ExilesI'm just poor so figured try to repair
07:10:46angelwolf71885i prefer to think f them as disabled brken implys beyond repair xD
07:11:35angelwolf71885lol i know what you mean
07:11:50krazykitnone of this has anything to do with rockbox. please stay on-topic
07:11:59angelwolf71885if you can fix it for the same or less cost as a new one but you learn while ding it why the hell nt xD
07:12:12Exilesmy original question is 100% about rockbox
07:13:23angelwolf71885did yisent there a way to probe the memory commands as they happen on the ipod via FW or 30 pin?
07:13:54angelwolf71885if so shuld be easy to wrte a driver with the memry command map
07:15:12Exilesthe only thing I notice is the debug reg is that port A changes from 7f to ff
07:15:58angelwolf71885try pressing the buttens n the remote and see if the memory map changes
07:16:26angelwolf71885isent there a way to see what commands are being invked or am i thinking of a later version?
07:16:28 Quit HellDragon (Read error: 104 (Connection reset by peer))
07:16:40 Join HellDragon [0] (
07:19:27Exilesonly change i get is gpio states a: goes from 7f to ff button presses on the remote do nothing yet
07:20:14Exileshaven't found a way to see the memory contents yet
07:20:16angelwolf71885dse RB see a deivice connected?
07:21:08Exilesnvm, rockbox dosn't see an acesory connecten
07:21:15Exilesit's always at 0
07:22:13angelwolf71885did that version of the mini have a 30 pin?
07:22:43angelwolf71885if s if you have a 30 pn AAP try pluging it in
07:23:13Exilesit does, and i dont
07:24:32angelwolf71885wait we knw the debug adress fr the mini port
07:25:00angelwolf71885i think the dck driver just needs t be writen to also check that adress
07:25:52 Part toffe82
07:26:08ExilesI can monitor the rx and tx line of the top port easily
07:26:10angelwolf71885now we just needs somone wh knows about the driver xD
07:27:02Exilesbreak out 2 ttl to usb serial adaptors, and monitor both pins
07:27:21angelwolf71885k lol
07:27:32angelwolf71885do you still have apples FW on the mini?
07:28:00angelwolf71885dose it reconice the remote?
07:29:27Exilesgotta add some songs via itunes to find out
07:29:37angelwolf71885k lol
07:30:41ExilesI don't really like itunes, aka the reason why I use rockbox
07:30:58angelwolf71885mac r pc?
07:31:11 Join Horschti [0] (n=Horscht2@xbmc/user/horscht)
07:31:16angelwolf71885can you live with winamp?
07:32:12Exilesmeh rather control it all myself
07:32:21 Quit advcomp2019 (Read error: 104 (Connection reset by peer))
07:32:27angelwolf71885not sure on ML_IPODS RB support
07:33:33angelwolf71885on my nan 1 g things didnt go well with RB and ml ipod
07:33:57angelwolf71885had iussues draging and droping and database rebuilding
07:34:06angelwolf71885wrked great when it was empty
07:34:23Lloreanangelwolf71885: Can you try to use whole words. That means "are" and not "r" and so on. This is in the channel guidelines. Take your time typing, and make sure to type things translation software can understand. Don't intentionally leave out letters.
07:34:35angelwolf71885but when i went to add more media drag and drop rebuilding dident work to see more media
07:34:52angelwolf71885i have issues with my O butten
07:35:19angelwolf71885most of the time im typeing OR
07:36:01Lloreanangelwolf71885: Just saying that doesn't solve issues for translation software when you type "butten" or "wrked"
07:36:38angelwolf71885im not gunna go trought it all just because f an O butten issue
07:37:55Lloreanangelwolf71885: You seem to be able to type O just fine for "lol" and plenty of other things
07:38:06LloreanPlease, skip the "lols" and use it where it's necessary in real words.
07:38:16Unhelpfulsaratoga: your point is made, just replacing floor_log2 with a non-braindead implementation saves .24MHz on e200... and i'm still not certain it's the *best* implementation, and i've not replaced another similar function.
07:38:24 Quit Horscht (Read error: 60 (Operation timed out))
07:38:31angelwolf71885its bent sometime it wors sometimes it misses
07:38:43LloreanWell you missed a k there.
07:38:54LloreanPlease, just take your time. There's no rush. Read what you're typing and correct it.
07:39:00angelwolf71885if you look careful some Os are missing bu latter in the string there is an o or tow
07:39:38LloreanIt's not just the letter o. Please, try to use real English and take your time.
07:41:22angelwolf71885i dont have to offer any explinashion
07:41:24Unhelpfula bad 'o' key hardly explains "bu", "latter", "tow"
07:41:52Lloreanangelwolf71885: The channel has posted guidelines. You're being asked politely to follow them.
07:42:13LloreanYou don't have to offer an explanation, no. But you are being told that we expect people using this channel to make a basic effort to use real English.
07:42:16Unhelpfulangelwolf71885: well, actually, there are rules for a reason. some users and developers may be reading channel logs through a machine translator. it doesn't know that you meant "later" instead of "latter", or what "bu" means.
07:42:51Unhelpfulso please, be considerate of others and take the time to do things properly.
07:43:01topikFlynDice: are the 5 minutes up yet?
07:43:18angelwolf71885thers not much i can do if one letter is haveing issues its likely more are starting to go
07:43:53 Quit panni_ ("( :: NoNameScript 3.81 :: )")
07:44:03FlynDicetopik: er yeah see [22:03] <FlynDice> topik: Could you try this and see if it helps thought you were gone...
07:44:22topikoh sorry
07:46:57topik"patch unexpectedly ends in middle of line"
07:47:29angelwolf71885hey Exiles you still there?
07:49:50angelwolf71885if the remote works in the apple FW then i suggest useing your sniffer to see what info you can gather
07:51:06angelwolf71885im out
07:51:09 Quit angelwolf71885 ("ChatZilla 0.9.85 [Firefox 3.5.5/20091102152451]")
07:51:12topikcompiling now
07:51:32 Join dmb [0] (n=Dmb@unaffiliated/dmb)
07:52:49Exileslooking for my usb sync cables atm, when i find it and find out it works I'll compile all the info I can on the serial side
07:59:09 Join AndyI [0] (n=pasha_in@
08:00:11***Saving seen data "./dancer.seen"
08:01:11 Join voltagex [0] (
08:05:19topikFlynDice: my uSD card is still 25Mbit/s
08:06:28FlynDicetopik: not what I was hoping.... did it fire up right away?
08:07:53topikit wasn't a cold start as i had to update rockbox first
08:07:59topikwill try again in a couple of minutes
08:08:33FlynDiceok, Ill be around for 30 mins or so then it's bedtime here
08:08:53topiki'll try not to stretch 5 minutes this time
08:08:53FlynDicethanks for trying
08:09:02FlynDice ;-)
08:11:17 Quit AndyIL (Read error: 110 (Connection timed out))
08:12:20topikit worked this time
08:12:28topikstarted i mean, but still at 25
08:13:04topikdo rev 7.0 and prod 4/2008 give you any idea if it's a troublesome old card?
08:33:45FlynDicetopik: No, I'm really interested in if the card got switched to HS timings(which would show 50.0 MBit/s). We've been running these cards at 62 MHz and they worked for some reason. Now that we've got them down to spec frequencies they seem to be acting up.
08:36:35FlynDicetopik: I think your card is just fine, I think our init sequence needs to be tweaked just a bit
08:36:45 Join Rob2223 [0] (
08:37:08 Join n1s [0] (n=n1s@rockbox/developer/n1s)
08:38:24topikit does seem to start normally now
08:38:36topikbut the card never gets anything but 25Mbit/s
08:39:32topikif you dream up any more patches, i'd be happy to try them :)
08:39:37topikthanks for the effort
08:41:28FlynDicetopik: Have you followed the forum messages about this problem?
08:43:20topiki have read them yes
08:43:41topikseems like the same thing i am experiencing
08:44:07topikat 62Mhz the sdhc cards run at 50Mbit/s ?
08:45:18n1sUnhelpful: yeah i saw that situation with the log functions, but if you are workign on it now i'll leave this to you :)
08:46:34Unhelpfuln1s: basically they're about the worse possible log2(x). ;)
08:47:18Unhelpfulsince floor_log2 is never passed a 0, we can actually use an even smaller table on coldfire than we use on arm
08:48:55Unhelpfuland we can save the x ^= x >> 1 step ;)
08:49:52Unhelpfuli brute-force-searched for a number that has the same property as a debruijn sequence, but with regard to multiplying by (2<<n)-1 instead of shifting <<n
08:50:30n1shmm, someone broke the manuals
08:50:49n1sUnhelpful: wow
08:51:07n1syou do take things to the extreme
08:51:20n1swhich is a good thing :)
08:52:05Unhelpfuln1s: it wasn't hard... and the coldfire has a decent multiplier and extension word immediates, so it can do a "real" multiply
08:54:23 Quit Rob2222 (Read error: 110 (Connection timed out))
09:00:25n1sB4gder: lots of manuals are missing from the website, they build fine localy though
09:06:33 Join tomers [0] (
09:06:45 Quit gevaerts (Nick collision from services.)
09:06:57 Join gevaerts [0] (n=fg@rockbox/developer/gevaerts)
09:09:36 Join flydutch [0] (
09:10:38FlynDicetopik: No at 62MHz the cards run at 62Mbits/sec.... The spec limit is 25 MHz for a standard speed SD card and 50 MHz for a HS SD card. We couldn't get them to run within the spec before though, rockbox wouldn't boot.
09:11:10 Join stoffel [0] (
09:13:57FlynDicetopik: We've still got the internals running slightly fast at 31MHz because If I take them down to 15.5 MHz(next lower option) I get data crc failures on writes to the card :/
09:14:20n1saaaaah, adding a splash "fixes" this crash!
09:15:32topikmake sense FlynDice. the goal of running them in-spec-slow is better battery life?
09:18:08FlynDicetopik: I don't think battery life is going to be meaningfully impacted. It's mostly just to bring it within spec. The OF runs the uSD at 16 MHz.
09:18:51topikso more a matter of principle than a technical requirement
09:19:24FlynDiceyes, Thanks for the help, bedtime!
09:23:09 Quit CaptainKewl (Read error: 110 (Connection timed out))
09:23:15 Quit B4gder (Read error: 110 (Connection timed out))
09:24:36 Join robin0800 [0] (
09:25:21 Quit robin0800 (Remote closed the connection)
09:25:41 Join robin0800 [0] (
09:29:43n1sok, this crash is weird and seems timing dependant
09:30:40n1sadding a call to pcm_play_stop() helps if i also add a sleep in pcm_play_stop, not if i sleep before calling it or after it returns
09:31:03n1son my h300 all is fine even without stopping playback
09:31:21n1sit "just works" as it were
09:37:24*gibbon_ still wonders if it would be feasible, to implement playable config files...
09:38:48n1swhat would happen when you play a config then?
09:39:11gibbon_i would like to see playlist-adjustable sound settings
09:39:38gibbon_for example "overlay" the pitch using a config file, mentioned in the playlist or load special EQ settings from within a playlist
09:40:36gibbon_i thought to mention a according config file in the playlist would be the easiest way to acquire that
09:40:43n1saha, probably not *that* hard to hack together, although the settings code might need some more smarts for it to be smooth
09:41:23gibbon_yes, i had second thoughts about how to handle these "overlay" settings in the normal sound settings
09:41:50gibbon_for example, what should the normal EQ adjustment show or do, if such a config file was loaded?
09:42:03gibbon_should that settings be saved? Probably not.
09:42:25 Join advcomp2019 [0] (n=advcomp2@unaffiliated/advcomp2019)
09:42:38gibbon_i will dive into the settings code and try to hack something together
09:43:40gibbon_even though i don't know where to implement it... would that be a codec? should the player "app" handle it? ...
09:45:21n1shmm, if you don't want to just mess with the regular settins it will be a bit harder i think, how you implement it is up to you but i wouldn't make a codec
09:46:55gibbon_i also thought so... a codec should leave its hands out of everything besides data input and audio output
09:47:39gibbon_i will think some more about it... but thanks for commenting
09:48:01gibbon_at least it seems, its not a particulary awful idea ;)
09:48:58 Quit advcomp2019 ("Leaving")
09:50:50 Join advcomp2019 [0] (n=advcomp2@unaffiliated/advcomp2019)
09:57:11 Quit n1s ("Lämnar")
09:59:50 Join n1s [0] (n=n1s@rockbox/developer/n1s)
10:00:16***Saving seen data "./dancer.seen"
10:01:31 Quit TopyMobile (Read error: 60 (Operation timed out))
10:08:26 Join TopyMobile [0] (
10:18:02 Quit TopyMobile (Read error: 60 (Operation timed out))
10:18:55 Join Tomis [0] (n=Tomis@
10:20:02 Quit Exiles (Read error: 104 (Connection reset by peer))
10:23:21n1sok, so i think that somehow the waiting for dma to die in the pcm-pp driver is broken and playback interrupts get fired when trying to record resulting in badness, the fact that the sleep delay only helps when inside of the pcm_play_(un)lock calls in pcm_play_stop seem to support this as the DMA interrupts are masked when the lock is held so any ongoing transfers will finish
10:29:06n1sanyone who knows pp and or pcm code who can comment?
10:31:55 Quit tomers ("ChatZilla 0.9.85 [Firefox 3.5.5/20091109125225]")
10:32:25 Join TopyMobile [0] (
10:33:20 Quit stoffel (Remote closed the connection)
10:35:32 Quit Tomis ()
10:38:58 Quit jon-kha (
10:38:58 Quit Slasheri (
10:53:54n1swow, the pith detector actually works while playback is running on h300...
10:54:01n1spitch, even
10:54:20n1sit doesn't stop playback
10:54:42n1si am surprised
11:13:22 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
11:20:14n1sis playback and recording dma supposed to work at the same time on pp?
11:20:47*n1s wonders where jhMikeS has gone
11:26:42 Quit phanboy4 ("Leaving")
11:35:47 Join stripwax [0] (
11:44:20 Join mc2739_ [0] (n=mc2739@rockbox/developer/mc2739)
11:47:20 Join pamaury [0] (
11:54:06 Nick Horschti is now known as Horscht (n=Horscht2@xbmc/user/horscht)
11:57:37 Quit mc2739 (Read error: 110 (Connection timed out))
11:57:54 Join geertvdijk [0] (
11:58:59 Quit geertvdijk (Client Quit)
12:00:18***Saving seen data "./dancer.seen"
12:01:13n1sugh, function allocating 1600byte array on stack *and* can call itself (through a few other functions) recursively :)
12:02:46 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
12:02:51 Quit TopyMobile ("Ex-Chat")
12:03:57 Join FOAD_ [0] (
12:05:55n1sthis thing should be rewritten to use the filebrowser directly
12:09:24n1sJdGordon: can a function used in a MENUITEM_FUNCTION return some magic value to make the menu exit?
12:16:32 Quit FOAD (Read error: 110 (Connection timed out))
12:16:33 Nick FOAD_ is now known as FOAD (
12:17:39 Join robin0800_ [0] (
12:19:51 Join MethoS- [0] (n=clemens@
12:20:53 Quit MethoS- (Remote closed the connection)
12:20:54 Join bnijk_ [0] (n=ush@unaffiliated/octopuswitharms)
12:21:05bnijk_i have an ipod 4g, what do i need to do to rebuild the database in rockbox
12:22:14gibbon_bnijk_: keep select pressed while on the database menu entry should fire up some database management menu
12:22:27 Join MethoS- [0] (n=clemens@
12:23:44bnijk_it does
12:23:49bnijk_Artists, Songs, etc
12:23:54bnijk_and then "Custom View" at the bottom
12:23:57gibbon_no... no
12:24:20gibbon_how long did you hold the select button?
12:24:49bnijk_ah i see
12:24:57bnijk_(this beats the regular ipod os)
12:25:01 Quit robin0800 (Read error: 60 (Operation timed out))
12:25:09markunbnijk_: we do our best :)
12:25:16gibbon_of cause it does... and many other original firmwares
12:25:21gibbon_you are welcome
12:40:01 Quit yosafbridge ("Coyote finally caught me")
12:40:12 Join yosafbridge [0] (
12:40:12 Join domonoky [0] (n=Domonoky@rockbox/developer/domonoky)
12:41:45 Join teru [0] (
12:59:03 Quit robin0800_ (Remote closed the connection)
13:00:39 Join petur [0] (n=peter@rockbox/developer/petur)
13:01:34 Quit bmbl (Read error: 110 (Connection timed out))
13:07:22 Nick fxb__ is now known as fxb (n=felixbru@
13:08:44stripwaxsaw an interesting post in vorbis-dev archives from 2005 - "The specific MDCT transform that Vorbis uses is not considered especially stellar [...] I'd hate for someone to spend six months writing an assembly version that's 50% faster, when simply switching to another algorithm would be 100% faster". I might add that to the fasterMDCT wiki page :)
13:22:33n1syay, i got the menu system to do what i wanted
13:25:27 Quit BHSPitMonkey (Read error: 60 (Operation timed out))
13:26:23n1snope, doesn't work right, the menu stuff is so weird
13:28:08n1sstripwax: does it mention which algo would be 100%faster?
13:30:45 Quit tarbo (SendQ exceeded)
13:31:37 Join tarbo [0] (n=me@unaffiliated/tarbo)
13:41:08 Join Slasheri [0] (i=miipekk@rockbox/developer/Slasheri)
13:47:56CIA-6New commit by teru (r23851): plasma: clean up button definition.
13:51:39 Join robin0800 [0] (
14:00:22***Saving seen data "./dancer.seen"
14:00:42n1sUnhelpful: are you currently messing with the log stuff, i just realised several of our codecs use the identical av_log2 code from ffmpeg, and thought it might be nice to unify that in the plugin lib
14:02:28n1swell s/several/ wma, cook and flac/
14:02:57stripwaxn1s - no .. (although from context, I'd intepret that to mean that DFT-based 'might' be better). Here's the thread anyway.
14:04:24mtstripwax: (a belated) pong :)
14:05:31n1sstripwax: ah, i guess Monty has some credibility when it comes to vorbis :) although he doesn't seem very sure
14:06:25stripwaxNow, apparently there was a vorbis (in fact, tremor-lowmem) patch to use n/4 ffts for the mdct - - but the patch link is very dead, and not even available at
14:07:35n1sstripwax: i think i have that patch
14:08:05stripwaxnp, just found a mirror -
14:08:27 Quit TaZzZ ("ZNC -")
14:09:39markunI wonder if they copied it from my server, since I had given it the same name :)
14:09:43stripwaxalthough (annoyingly) rockbox doesn't use the lowmem / half-mdct version of tremor so it's not directly applicable
14:09:48stripwaxmarkun - :-)
14:10:05markunstripwax: I also have a copy of the thesis
14:11:05markunah, they have that one as well :)
14:11:27stripwaxmt - so, I noticed on the FasterMDCT page, you/saratoga mentioned some analysis had already been done on some of the various alternative mdct implementation out there - can that be added to the wiki?
14:11:28markun(with again the same rename as mine)
14:12:21mtstripwax: iirc saratoga has the numbers.
14:13:42 Quit flydutch ("/* empty */")
14:20:14 Join TaZzZ [0] (
14:20:39stripwaxit's funny, every google hit I've seen regarding some vorbis/mdct research seems to have saratogas name on the page somewhere :)
14:21:04n1smt: could you provide a (few) test file(s) for cook to be included in the test_files set, it's very nice for benchmarking and testing without having to track down an encoder and makes for easy comparisons (they are all made from the same file)
14:22:42stripwaxmt - is there already a patch for changing rockbox tremor to use a half-mdct?
14:23:18CIA-6New commit by teru (r23852): boomshine: use theme colour to draw text so that text would be more readable with white background.
14:23:28mtn1s: ok.
14:23:37n1smt great :)
14:23:48 Join MegaTron [0] (
14:24:10mtstripwax: I'm not really aware of a patch that does that.
14:24:14n1smt: you can use one of the lossless samples in as a source
14:26:14 Part MegaTron
14:30:24 Quit elcan (Remote closed the connection)
14:30:39 Join elcan [0] (i=user36@
14:34:07Unhelpfuln1s: well, part of the trouble is going to be that there are several things using integer log2, each with different expectations for handling of 0 inputs
14:34:47n1sUnhelpful: yeah, i know, flac, wma and cook does however use the *exact* same function
14:36:39Unhelpfuln1s: iirc the ffmpeg log2 does two stages of binary search, and then throws the last 8 bits to an LUT?
14:36:40n1si came up with this for log(0) == 0 with the clz instruction: int log2 (int v) { int lz = __builtin_clz(v); return 31 - lz + (lz >> 5)); // log2(0) == 0
14:37:04 Quit elcan (Remote closed the connection)
14:37:09 Join elcan [0] (
14:37:10n1sUnhelpful: yes
14:37:34Unhelpfuln1s: keep it simple, (__builtin_clz(x) & 31) ^ 31 ;)
14:39:06n1smy bit fiddling foo is clearly not at your level
14:40:25Unhelpfulit's also pretty easy to spit out something that you came up with a few days ago. i got to looking at clz and log2 again a week or so ago when i had a good look at ffs-arm.S
14:40:28 Quit pamaury (Read error: 113 (No route to host))
14:41:57Unhelpfulunfortunately my bit-fu is not up to proving whether or not any additional numbers are constructible with shifted add/subtract instructions if i use sub as well as rsb/add. will have to tweak my brute-forcing program to do it. ;)
14:44:12n1snothing wrong with building tools to do what you can't :)
14:46:42Unhelpfulmaybe, but better methods usually win over more brute strength - as in the case of directly generating sequences of instructions to construct values, and testing constructible values. threw *way* too much CPU time at brute-force searching the integers for the right numbers, and *then* calling gcc to see how it decided to construct them
14:54:42 Join funman [0] (n=fun@rockbox/developer/funman)
14:56:36stripwaxis log2 used enough inside inner loops to really warrant hand optimsation?
14:56:50Unhelpfulalso my bit-fu is weak, or i'm just tired. you want to swap the ^ and &, as the & is where you collapse from 33 to 32 possible outputs
14:57:29Unhelpfulstripwax: there are two different log2-ish functions in faad. improving one of them saved .24MHz on e200.
14:57:37n1sstripwax: using the clz instruction gave 12% speedup on flac decoding on the beast
14:59:04Unhelpfuland the gain will be larger on beast as it can produce the correct values in 2-3 ops :)
14:59:22Unhelpfulalso faad's log2 logic was especially braindead :)
15:00:38mtn1s: How do I upload the files there ?
15:00:44 Join arohtar [0] (n=faemir@
15:01:24n1smt: i guess you have to ask Bagder or Zagor, they don't seem to be around now though
15:01:53mtguess I'll just e-mail one of them then.
15:02:17n1sdo that, do you have the files uploaded somewhere?
15:03:53mtnot yet, but I can upload them.
15:04:10stripwaxn1s (wow)
15:04:31mtI'll host them for bagder/zagor to be able to download them anyway.
15:04:37n1smt: i guess that would be easiest so they can grab them when they have time, (also i'd like to run a few benchmarks now :))
15:05:02mtOh, didn't know that ! :)
15:08:11CIA-6New commit by teru (r23853): spacerocks: Fix indentation and remove trailing spaces. Fix some minor bugs.
15:16:55 Quit liar (Read error: 113 (No route to host))
15:16:55 Join jon-kha [0] (
15:17:22 Quit faemir (Read error: 110 (Connection timed out))
15:19:15 Join Strife89 [0] (
15:27:20 Join fdinel [0] (
15:34:50CIA-6New commit by rmenes (r23854): A few more GoGear SA9200 plugin keymaps.
15:42:32 Join etaOh [0] (n=eta2h@
15:42:53etaOhGood morning
15:44:47etaOhRunning RB on a Sansa e2800v2 - really like it a lot, just wanted to say thanks to all the devs
15:48:21etaOhIs it possible to have the device not boot the original firmware when plugging in the USB?
15:48:34funmanetaOh: no
15:49:27 Nick fxb is now known as fxb__ (n=felixbru@
15:50:32etaOhok, thanks
15:51:03 Join pamaury [0] (
15:51:36mtfunman: Do you have any idea when is sandisk releasing a firmware update for the clip+ ? I found a solution for reading the raw data off of the flash, but it might not be worth the effort if the upgrade is soon.
15:53:33funmanmt: no, i'm checking regularly the forum to see if it appeared
15:56:35mthmm .. well, I found a website, explaining how to attach an SD card reader to a flash, and dump an image of the raw data with it. The only problem is that I neither have the tools, nor the expertise, to solder smd, so I'll need to find someone who can do that first.
15:57:24gevaertsetaOh: hold the appropriate key (I think select)
15:58:32funmanetaOh: ah, you can plug USB to charge only (no data connection) by holding select while plugging
15:59:01funmanmt: there's no hurry, you can study the Clipv2 code to reverse engineer SD while waiting ;)
16:00:15mtuhm .. will look into that. :)
16:00:25***Saving seen data "./dancer.seen"
16:01:21 Join stoffel [0] (
16:01:52etaOhok cool, so it's possible to listen to music while charging then
16:02:26etaOhThis little mp3 player was the best $80 I've spent in a long time :)
16:08:21 Join GeekShado_ [0] (
16:12:26 Quit teru ("Quit")
16:12:44etaOhok, well thanks again, for what it's worth your hard work is much appreciated - ttyl
16:12:57 Quit etaOh ("Ex-Chat")
16:15:43funmanmt: i'd be happy to help with this SD code, I had worked a bit on it but now i lack motivation for working alone
16:18:00mtfunman: That'd be great. Given my unprecedented experience in SD, I don't think I can do much alone. :)
16:19:50funmanmt: first thing is to start disassembling the Clipv2 OF
16:20:10funmani'm using the v.16 (last is v.32)
16:23:25 Quit GeekShadow (Read error: 110 (Connection timed out))
16:23:27 Join said [0] (
16:32:23 Quit MethoS- (Remote closed the connection)
16:34:21 Quit arohtar (Read error: 104 (Connection reset by peer))
16:47:25 Quit said (Remote closed the connection)
16:53:54 Join hebz0rl [0] (
16:54:07CIA-6New commit by tomers (r23855): On RTL mode, flip + and - buttons (Cowon D2)
16:54:27stripwaxmt - ... so the problem with half-mdct is really that you can't do the windowing in-place - so you save some memory by needing to store half blocks initially, but you need to either then unroll elsewhere or do the windowing into a separate buffer. UNLESS the windowing step happens at the exact same time as pcmbuf_insert (which probably means inlined into dsp_process) - not ideal
16:57:34 Join robin0800_ [0] (
16:57:42 Join B4gder [241] (n=daniel@rockbox/developer/bagder)
17:00:29 Join tomers [0] (
17:03:09n1sok, using a clz based log2 for wma gives a *tiny* speedup on the beast ~.2%, no difference on cook and the extra handling of log(0) slows down flac by 2%
17:04:55CIA-6New commit by tomers (r23856): Cowon D2: Fix standard keymaps ...
17:10:45stripwaxn1s - yeah I didn't think log2 was used much by the other codecs
17:11:07stripwax(where much = somewhere in an inner loop that would make much difference)
17:11:42n1sstripwax: apparently faad used it a bit, unhelpful got some promising results with his amazing asm function for armv4
17:13:52n1sanyway, unifying code is nice
17:14:04 Quit robin0800 (Read error: 104 (Connection reset by peer))
17:21:35 Join paulk_ [0] (
17:21:54paulk_hello! Is it possible to use <time.h> with the rockbox api ?
17:24:54paulk_when I write #include <time.h>, it doesn't work
17:25:47n1sno, we don't have <time.h>
17:26:11paulk_can't do randoms ?
17:26:18n1ssure can
17:26:22paulk_how ?
17:26:30 Quit Strife89 (Read error: 104 (Connection reset by peer))
17:26:32gevaertspaulk_: is this for a plugin?
17:26:44paulk_yes :)
17:26:48 Join Strife89 [0] (
17:26:53paulk_a game, but no very interesting
17:27:01gevaertsn1s: we do have time.h in firmware/include
17:27:19gevaertsmight not be usable from plugins though
17:27:27*gevaerts isn't too familiar with plugins
17:27:52paulk_I'll try
17:27:55n1spaulk_: the best way for you to get an answer is to tell us what you want to do
17:28:28paulk_n1s, I want to generate a random number between 0 and 100
17:29:26n1spaulk_: we have two functions for that in the plugin api, srand to seed the randomizer and rand to generate a random number
17:29:49paulk_lenombremystere.c:(.text+0x54): undefined reference to `srand'
17:29:55paulk_lenombremystere.c:(.text+0x58): undefined reference to `rand'
17:30:05n1sso for example srand(current_tick); int v = rand() %100;
17:30:15n1spaulk_: are you calling them throught the api pointer?
17:30:25paulk_oh no, sorry :(
17:30:29 Join Strife1989 [0] (
17:30:56paulk_ir works
17:31:12paulk_now I need to get "current_tick"
17:31:21paulk_(typicaly with time.h)
17:31:58n1sah, sorry that is also in the plugin api so do rb->srand(rb->current_tick);
17:32:45paulk_It seems to work :)
17:33:16*gevaerts thinks he can guess what the game is about, but maybe he should guess what the number is instead :)
17:33:21 Join faemir [0] (n=faemir@
17:33:53paulk_gevaerts, :D exactly
17:34:35paulk_I've made it in C/GTK, C/Bash and now in C/Rockbox
17:38:10 Quit faemir (Read error: 104 (Connection reset by peer))
17:38:20 Join Rondom [0] (
17:40:07 Nick YPSY is now known as Ypsy (
17:41:49 Join faemir [0] (n=faemir@
17:42:03 Join flydutch [0] (
17:42:35CIA-6New commit by tomers (r23857): Touchscreen keymapping enhancements ...
17:46:48 Quit Strife89 (Read error: 110 (Connection timed out))
17:47:45CIA-6New commit by nls (r23858): Move av_log2 function and asociated table to the codec lib, remove 3 identical implementations, always have LUT in iram (gives a *tiny* speedup on ...
17:52:21 Join panni_ [0] (
17:54:00 Join liar [0] (n=liar@
18:00:27***Saving seen data "./dancer.seen"
18:09:00 Join Strife89 [0] (
18:16:11 Join Kitr88 [0] (
18:23:33 Quit Kitr88 (Read error: 60 (Operation timed out))
18:23:40 Join Kitr88 [0] (
18:25:44 Quit Strife1989 (Read error: 110 (Connection timed out))
18:25:52 Join Highlander [0] (
18:32:40 Join Tomis [0] (n=Tomis@
18:33:15 Join Strife1989 [0] (
18:34:26 Quit Kitar|st (Connection timed out)
18:36:54 Nick mc2739_ is now known as mc2739 (n=mc2739@rockbox/developer/mc2739)
18:41:47 Join phanboy4 [0] (
18:42:44 Quit liar (No route to host)
18:43:53 Join liar [0] (n=liar@
18:44:21 Quit stoffel (Remote closed the connection)
18:45:18CIA-6New commit by tomers (r23859): Cowon D2: In WPS context, assign BUTTON_POWER to ACTION_WPS_BROWSE
18:45:27 Quit robin0800_ (Read error: 54 (Connection reset by peer))
18:48:41CIA-6New commit by tomers (r23860): Touchscreen button mapping enhancement in WPS context ...
18:49:30 Quit Strife89 (Connection timed out)
18:50:14 Quit Tomis (Read error: 54 (Connection reset by peer))
18:50:39 Join Tomis [0] (n=Tomis@
18:54:58 Join dfkt_ [0] (i=dfkt@unaffiliated/dfkt)
18:55:00 Quit dfkt_ (Remote closed the connection)
18:56:03 Nick Strife1989 is now known as Strife89 (
19:00:10 Join klapaucjusz [0] (
19:00:31klapaucjuszI have a patch that works fine in the simulator, but crashes my iPod.
19:00:53klapaucjuszFS #10832
19:01:07klapaucjuszAny hints on debugging on the real hardware?
19:01:44 Quit dfkt (Read error: 60 (Operation timed out))
19:02:19funmanklapaucjusz: how does the ipod crash ? (show data abort ?)
19:02:45klapaucjuszIt gives me a white background with the message "division by zero at whatever"
19:02:59klapaucjuszWhen I reboot, the flash is erased.
19:03:04funmanload the rockbox.elf file in a disassembler and check what is at this address
19:03:26klapaucjuszHmm, yep, I guess I could reflash my iPod once again ;-(
19:03:28 Quit liar (Read error: 113 (No route to host))
19:03:38klapaucjuszBut it won't do me much good if the crash is in the codec, will it?
19:04:06 Join Tomis2 [0] (n=Tomis@
19:05:28klapaucjuszBy the way −− I've found that the iPod (nano 2g) erases the flash on every rockbox crash.
19:05:44klapaucjuszSomething wrong with the exception handler?
19:06:08stripwaxklapaucjusz - is your patch same/similar as FS #10160?
19:06:10 Quit Tomis (Read error: 60 (Operation timed out))
19:06:10 Nick Tomis2 is now known as Tomis (n=Tomis@
19:07:18stripwaxklapaucjusz - (the FS# I mentioned is also for mdat chunks in mp4/m4a...)
19:07:20klapaucjuszstripwax: gosh no.
19:07:24klapaucjuszMine actually works.
19:07:39stripwaxit sounds like it crashes your ipod, no? :)
19:07:45 Quit Highlander ("Quitte")
19:07:52klapaucjusz(1) FS #10160 only deals with the metadata parser, while mine does both the parser and the codec.
19:08:12stripwaxyeah - definitely need both
19:08:14klapaucjusz(2) FS #10160 goes through the whole file, while mine only goes through the file if it's actually unseekable.
19:08:38klapaucjuszi.e. the behaviour is unchanged in the case of properly encoded MP4s, there's no seeking backwards in that case.
19:09:13klapaucjusz(3) FS #10160 adds code to deal with chunks over 4GB, which mine doesn't attempt to deal with. (It's an unrelated issue.)
19:09:34klapaucjuszstripwax: point taken.
19:09:39klapaucjuszI'll correct myself, then.
19:09:48klapaucjuszMine is actually supposed to work ;-)
19:10:02 Quit Tomis (Read error: 54 (Connection reset by peer))
19:10:56stripwaxklapaucjusz - does it work successfully on other targets (just crashing ipod nano 2g)?
19:11:22 Join Tomis [0] (n=Tomis@
19:11:22klapaucjuszNo idea.
19:11:25domonokyklapaucjusz: you can also use the .map file to findout where this crash happens.
19:12:07klapaucjuszdomonoky: excellent
19:12:32klapaucjuszdomonoky: er, no.
19:12:41klapaucjuszIt happens above the top address listed in that file.
19:13:16domonokydoes the crash happen in the codec or the main binary ?
19:13:27klapaucjusz(Audiobuffer is 81729f8, and while I didn't write down the address, it definitely started with a 9.)
19:13:32klapaucjuszdomonoky: how to I find out?
19:13:55 Join liar [0] (n=liar@
19:14:38 Join CaptainKewl [0] (
19:15:04domonokythe address should tell you.. take look at the codec .map file..
19:15:32klapaucjuszAh, I see.
19:15:38klapaucjuszThere's a whole bunch of map files, then.
19:16:08 Join Tomis2 [0] (n=Tomis@
19:16:18klapaucjuszExcellent. Time to crash my iPod again, then.
19:16:28 Quit Tomis (Read error: 104 (Connection reset by peer))
19:16:29 Nick Tomis2 is now known as Tomis (n=Tomis@
19:16:53klapaucjuszI'll let you know how it goes.
19:16:57klapaucjuszUnrelated question.
19:17:05klapaucjuszHow do I request write access to the wiki?
19:17:28klapaucjuszAnd what are the rules for requesting commit access?
19:18:03klapaucjusz(I'm not asking for commit access, I'm just asking what the conditions are.)
19:18:23domonokyyou ask some one here, and give your wikiname for the wiki.
19:18:32klapaucjuszOkay, I want "jch".
19:19:11klapaucjuszAh, I see, I need to register myself. Hold on.
19:19:47domonokycommit access is "earned" by putting enough good enough patches into flayspray.. if we think you are "worth" it, you will be offered commit access :-)
19:20:13domonokyand "jch" is not a valid wiki account name.. :-)
19:21:29klapaucjuszI always use jch.
19:21:39klapaucjuszI feel my identity is being denied ;-)
19:21:55klapaucjuszRe commit access, that's reasonable enough.
19:23:32 Quit amiconn (Nick collision from services.)
19:23:36 Join amiconn_ [0] (i=quassel@rockbox/developer/amiconn)
19:23:43 Nick amiconn_ is now known as amiconn (i=quassel@rockbox/developer/amiconn)
19:25:56 Quit flydutch ("/* empty */")
19:26:03 Quit pixelma (Nick collision from services.)
19:26:05 Join pixelma_ [0] (i=quassel@rockbox/staff/pixelma)
19:26:25 Nick pixelma_ is now known as pixelma (i=quassel@rockbox/staff/pixelma)
19:28:44 Join reich [0] (n=autodele@
19:30:08reichhi does anybody know that ipod video manual can't be downloaded right now?
19:30:33reichor am i the only one who cant?
19:32:14klapaucjuszTrying on a different file, I cannot reproduce the crash
19:32:19klapaucjuszI'm not sure if it's :-( or :-)
19:32:27klapaucjuszI'll try the previous file...
19:32:53 Join Tomis2 [0] (n=Tomis@
19:32:55 Quit CaptainKewl (Read error: 60 (Operation timed out))
19:33:46reichok thx how did it work out?
19:36:50klapaucjuszOkay, I managed to reproduce the crash. It appears it's in get_sample_offset.
19:38:06 Quit paulk_ ("Ex-Chat")
19:38:06 Quit Tomis (Read error: 54 (Connection reset by peer))
19:38:48klapaucjuszOkay, it's unrelated.
19:38:57reichuhm? sorry what do u mean by get_sample_offset?
19:39:34klapaucjuszm4a.c line 204.
19:39:48klapaucjuszThere's a division.
19:39:58klapaucjuszFor this particular mp4 file, prev_chunk_samples is 0.
19:40:00klapaucjuszDon't ask me why.
19:40:21 Join saratoga_ [0] (i=463f90ed@gateway/web/freenode/x-xzlgriikeqcqunup)
19:42:03klapaucjuszI don't understand this code, since I don't have the mp4 specs. I don't understand what's going on in read_chunk_stsc, and my guess is that there's an off-by-one error in there.
19:42:13klapaucjuszFolks, where do I find info about the mp4 container format?
19:42:34funman14496 spec
19:42:52 Nick Tomis2 is now known as Tomis (n=Tomis@
19:43:18saratoga_klapaucjusz: i'll see if I can find you some docs today
19:43:28saratoga_is your tracker email address ok
19:43:41CIA-6New commit by tomers (r23861): Touchscreen mapping: Fix setting context ...
19:44:09klapaucjuszsaratoga_: yes it is. Feel free to send directly to the address you'll find on
19:44:53saratoga_ok i'll look around, i forget which docs i have on my other computer
19:45:10 Quit Tomis (Read error: 104 (Connection reset by peer))
19:45:11 Join Tomis2 [0] (n=Tomis@
19:48:04klapaucjuszIt works!
19:48:56klapaucjuszdomonoky: thanks for telling me about the .map files.
19:52:58 Join Omlet [0] (i=omlet05@
19:53:27klapaucjuszFS #10833
19:54:31 Join Strife1989 [0] (
19:54:47saratoga_its not the mp4 standard, but it looks like it has some info about it
19:55:06saratoga_14496-14 is the actual mp4 file format, and i don't think i have access to it
19:55:12klapaucjuszExcellent, that's exactly what I'm looking for.
19:56:16saratoga_"This specification may be used as the basis as the specific file format for a restricted purpose: for example, the MP4 file format for MPEG-4 and the Motion JPEG 2000 file format are both derived from it."
19:56:25saratoga_so its not exactly the same
19:56:34klapaucjuszI see.
19:56:59 Quit Strife89 (Read error: 60 (Operation timed out))
19:57:33klapaucjuszBut it does describe the stsc chunk, which is the part where I suspect a bug.
20:00:30***Saving seen data "./dancer.seen"
20:03:08 Quit reich ()
20:03:45CIA-6New commit by tomers (r23862): Yes/No screen: Add proper message for touchscreen targets
20:10:02klapaucjuszMay I humbly request wiki modification rights for JuliuszCh?
20:11:50 Join Tomis [0] (n=Tomis@
20:12:59 Quit Tomis2 (Read error: 104 (Connection reset by peer))
20:13:14funmansaratoga_: ah I think part14 obsoletes part12
20:13:36saratoga_but i've never found a copy of part 14
20:13:48saratoga_just books about it
20:17:30klapaucjuszWhat's the story about multiple mdat chunks?
20:17:44klapaucjuszAre we supposed to concatenate them?
20:20:03 Nick Strife1989 is now known as Strife89 (
20:23:09CIA-6New commit by tomers (r23863): Updated Hebrew translation
20:25:19 Quit funman ("i still see you trolling")
20:31:19 Quit saratoga_ ("Page closed")
20:36:43 Quit Zarggg (Read error: 104 (Connection reset by peer))
20:37:34 Join Zarggg [0] (
20:39:47 Quit dmb ("Leaving")
20:43:31 Nick Ypsy is now known as YPSY (
20:43:41 Join domonoky1 [0] (
20:46:20 Join solexx [0] (
20:47:39 Quit domonoky (Read error: 60 (Operation timed out))
20:49:53 Quit Zarggg (Read error: 104 (Connection reset by peer))
20:50:12 Join Zarggg [0] (
20:50:15 Quit Zarggg (Read error: 104 (Connection reset by peer))
20:53:52 Quit Llorean ("Leaving.")
20:55:30CIA-6New commit by mcuelenaere (r23864): Ingenic Jz4740: fix CPU boosting (was reversed)
20:55:57 Quit Tomis (Read error: 60 (Operation timed out))
20:57:51 Quit solexx_ (Read error: 110 (Connection timed out))
21:00:58 Join DerPapst [0] (
21:02:50 Join Zarggg [0] (
21:02:59 Join CaptainKewl [0] (
21:07:23 Join ecaz [0] (
21:08:37 Join Casainho [0] (
21:09:25ecazHello friends! Am I clear in understanding that Rockbox is not stable on 7th Gen iPod Classic or is it completely unusable?
21:10:15bluebrotherecaz: it does not even run on any Ipod Classic
21:10:26stripwaxecaz - it doesn't even exist
21:10:35ecazOh dear
21:10:57bluebrotheror do you see it in any of the unstable or unusable categories?
21:11:57ecazI do not. I just assumed (oh dear!) that "iPod" included the classic
21:12:07ecazbut, that's what I get isn't it =)
21:17:27 Join arohtar [0] (n=faemir@
21:22:17Unhelpfuln1s: actually, i think i see a solution in 3 ops for armv6, using the saturation instructions...
21:24:48stripwaxecaz - the front page of specifically says ipod 1g thru 5.5g are supported; and "if your player is not listed above, it is not supported" .. sorry ..
21:25:28 Quit tomers ("ChatZilla 0.9.85 [Firefox 3.5.5/20091109125225]")
21:26:13stripwaxecaz - I believe some investigative work has taken place trying to break the 6th and 7th gen classic encryption (which prevents us from loading rockbox onto those players; and also prevents us understanding how the hardware works) - but I don't think anything substantial yet
21:27:32 Join jae [0] (
21:27:51jaeHola :D
21:28:20jaeJust installed RB on my Sansa Fuze (v1, of course)... worked like a charm
21:28:36jaeOnly... I can't play anything, and the manual seems to be missing
21:29:06jaeAnyone have an idea (or know, even) what happened to the manual, or why I can't see it (all I get is a 404)
21:29:54stripwaxjae - what does 'can't play anything' mean? any errors? files not showing up? you can try loading one of the manuals for other devices if it helps (to understand rockbox - won't tell you anything device specific for your fuze however)
21:32:08jaeAll I get is the ROCKbox logo at the top (with a battery indicator and a clock in the line above) and a version string at the bottom
21:32:36jaeWhichever button I press gives me... nothing, no reaction at all
21:32:55bnijk_yeah, why are all the manuals empty
21:33:44jaeAnd since I can't read the (temporarily, I presume) nonexistent manual, I don't know if that's what I should expect, if it's just initializing, or which button I have to press or whatever...
21:35:09 Quit faemir (Read error: 110 (Connection timed out))
21:35:12*bluebrother spots that the download server is also missing files
21:35:15mc2739jae: do you have a microsd card in your fuze?
21:35:39jaemc2739: yup, I do... that a problem?
21:35:58 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
21:36:26mc2739jae: Try removing it. There was a recent change that seems to be causing some problems with the microsd
21:36:57mc2739jae: you may be able to insert it after booting, but it may still cause problems
21:37:04jaemc2739: Ah... I see. W/o the card it works.
21:39:37 Quit Strife89 (Read error: 60 (Operation timed out))
21:42:34bluebrotherjae: built you a manual from current svn:
21:44:48jaebluebrother: *Thanks*!
21:45:27*jae notes to self: get sources... or find the ones downloaded sometime in the past :D
21:45:49*bluebrother wonders why the "Rockbox" on page 2 isn't centered when building with TeXlive on OS X
21:55:28 Quit hebz0rl ("Ex-Chat")
22:00:31***Saving seen data "./dancer.seen"
22:06:23CIA-6New commit by bluebrother (r23865): Fix title centering. ...
22:08:37 Quit Casainho (Read error: 113 (No route to host))
22:08:48 Quit bnijk_ (Read error: 104 (Connection reset by peer))
22:10:06 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
22:14:41 Join Strife89DS [0] (
22:19:11gevaertsZagor: "Server message: Fatal build error: Missing log file. You have been temporarily disabled."
22:22:09 Join p3tur [50] (n=petur@rockbox/developer/petur)
22:22:43mc2739FlynDice: I'm also seeing lockups on boot with my e260v2 whem the microsd card is inserted. Removing the card allows proper bootup.
22:23:24saratogastripwax, mt: I've updated the wiki with my mdct estimates, i'd be interested to know if you think they are reasonable
22:25:48ecazstripwax: is there more detailed information on the 'encryption' that ipods use?
22:26:44ecazI guess the firmware used to write to the device is hard coded and you don't have the pathways? (limited knowledge of Electronics on my end)
22:27:03stripwaxsaratoga - will take a look
22:29:17stripwaxecaz - the firmware is encrypted (= scrambled, so we can't read/understand it). naturally Apple haven't provided the password to unencrypt the firmware, which means we can't decrypt it to find out what it does / how it works, and we haven't yet found any 'backdoor' method of finding the decryption password ourselves
22:29:54ecazWe need DVD Jon!
22:30:07stripwaxthe ipod will only load encrypted firmware. if we can't encrypt rockbox in the same way, the ipod won't run it. plus, if we can't decrypt the Apple firmware, we don't know how to write device drivers that talk to the ipod hardware and make it work ..
22:30:18saratogawhich ipod is this?
22:30:31ecazhow were old devices done? Did they lack encryption?
22:30:37stripwaxmostly, yes.
22:30:38saratogayeah i think only the 6G and lower has been cracked
22:30:57stripwaxsaratoga - we don't have rockbox on 6g either though yet (or do we..?)
22:31:09stripwaxlikely, nobody has really taken a look at the 7g yet
22:31:14saratogai think its possible to launch a JTAG session and run code that way
22:31:28saratogabut IIRC linuxstb tried on the 6.5 or maybe the 7 and it didn't work
22:31:42saratogaask TheSeven or check the linux4nano wiki/logs and see
22:32:25saratogabasically i think the nano2g, nano4g and the first classics can be cracked
22:32:25CIA-6New commit by unhelpful (r23866): Slightly improve av_log2 in codeclib on ARMv6.
22:32:26ecazis rockbox in C?
22:32:36stripwaxecaz - and assembler
22:33:13ecazstraight C not ++
22:34:36saratogathe mul counts I posted seem a little low for 2048 point MDCT
22:36:01saratogabut i traced through the fft twice
22:36:49saratogai bet i'm not thinking about the overlap correctly
22:36:54stripwaxhrm, how many fixed point muls and fixed point adds does the tremor MDCT use
22:37:02saratogai'm not sure its hard to read
22:37:39stripwaxseems you're comparing real-life MHz to theoretical MHz for the two algos so really not like for like (e.g. neglecting all the pointer math, bitreverse, etc)
22:37:50saratogaof course
22:37:51stripwaxtrue, it is hard to read
22:38:07*jae is happy just for having rockbox on his Fuze, finally
22:38:46saratogabut assuming 1 load, 1 store per multiply and per add and no use of LDM on PP is pretty conservative
22:39:06saratogahmm maybe i should have assumed two loads per multiply
22:39:12stripwaxyeah, and two stores
22:39:12saratogawell whatever its just an estimate
22:39:19saratogatwo stores?
22:39:28stripwaxah, no, one store - was thinking of the complex mults.
22:39:58saratogai wonder if i'm off by a factor of two here because each time domain block has parts of two transform blocks
22:40:56stripwaxoh, the "84 clocks per sample" bit ?
22:41:04saratogastripwax: does the 2048 point transform in vorbis (the long block one) return 2048 or 4096 points?
22:41:10stripwaxdepends how you've defined 2048-point transfer - input points or output points?
22:41:15stripwaxhaha. right, exactly
22:41:20saratogayeah thats what i'm wondering
22:41:47saratogai thought it was 2048 in frequency, and 4096 in time
22:41:47stripwax(btw vorbis long blocks don't have to be 2048 .. can be more, could conceivable be less)
22:41:49 Quit bmbl ("Bye!")
22:41:53saratogaso a 2048 block takes care of 2048 frequency samples (out of 44100 per stereo channel)
22:44:35mc2739Unhelpful: yellows on the build table
22:45:34FlynDicemc2739: did you try the patch I posted on the forum by any chance?
22:45:52stripwaxwell, vwin2048 window generates 1024 output samples, from a pair of overlapping half-blocks. (note vwin2048 is only 1024 points as longblock-longblock window function is symmetrical)
22:45:55mc2739FlynDice: no, I have not tried it yet
22:46:20saratogaso i'm off by 2?
22:46:25stripwaxthink so
22:47:21FlynDicemc2739: would you try that when you get a chance, it seemed to help topik with his fuze, I think I've got a handle on this but I'm still trying to get info
22:47:52 Join Vampireking [0] (n=vampirek@
22:48:12mc2739FlynDice: ok, I'll give it a try it now
22:48:40saratogastripwax: so we're actually looking at about 16MHz with the unoptimized transform?
22:49:19stripwaxtheoretically? I think so..
22:50:02stripwaxbut I would not trust the "16MHz vs 16.79MHz" argument ...
22:50:06 Quit jae ("leaving")
22:50:45stripwaxtrying to work out how many muls+adds used for 2048 tremor right now
22:50:56 Join jae [0] (
22:51:26mtseems like I'm late !
22:51:56saratogalooking at the ffmpeg mdct, maybe i misunderstood the block sizes when i computed those totals
22:52:55saratogathe pre/post rotation uses N/4*(4MUL + 5 ADD) + N/8*(8MUL + 11 ADD)
22:53:06mtsaratoga: I don't understand something ..
22:53:12mtAh was just going to ask about that
22:53:16saratogabut I assumed N=2048 when it could be N=1024 for a large block
22:53:52CIA-6New commit by unhelpful (r23867): Fix yellow: mixed signed/unsigned in conditional.
22:54:01mtThe number of muls on the fasterMDCT page is the same as that you calculated for the fft only, right ?
22:54:14saratogamt: its also got the rotation counts added in
22:55:37mtOh, so those 2 lines contain the rotation counts ? I missed that :
22:55:38mt> /* 2048 pt = 512*(4MUL + 5 ADD) + 4328 MUL, 12219 ADD + 256*(8MUL 11
22:55:40mt> ADD) = (4096 MUL+ 5376 ADD) + FFT512 = 8424 MUL, 17595 ADD */
22:56:26saratoga4328 MUL, 12219 is supposed to be the fft count i think
22:56:31saratogabut i think its actually wrong
22:57:34saratogaoh no its correct
22:57:40saratogaor at least thats what my notes say
22:58:02saratogaof course with the optimizations we added, FFT8 now uses 4 less muls, and FFT16 8 less
22:58:07 Quit Strife89DS (Connection timed out)
22:58:19saratogaso we probably do 10-20% better
23:03:21stripwaxsaratoga - you're not including windowing, right, just the plain mdct here. i'll do same for tremor
23:04:18mtsaratoga: a shift takes one clock right ? (could be counted as an add i mean - for fixmuls?)
23:05:22stripwaxshift should probably be absorbed into add, as it can be done in zero clocks on arm usually
23:05:35stripwaxor absorbed into load/store (which you're not counting, hence 'for free')
23:06:15VampirekingHello i have a problem, i have moved some data on my sandisk sansa e260v2 from my microsd to the flash memory that was abortet with a sd transfer error. Now rockbox wont start and the original firmware says "Not enough space for music db please free up 90 mb" but the player wont detect on my pc anymore. I dont know how I can format my player now.
23:06:19stripwaxor are you including bitreverse, etc, in your stats?
23:06:43mtno, that's what I was looking for. Thanks :)
23:07:06saratogastripwax: I just counted adds/muls
23:07:06saratogai didn't do windowing
23:08:11saratogabit reversal is pretty fast since its folded into rotation, but i didn't look at it
23:09:04saratogaactually i think bit reversal should be free aside from 1 cycle to compute the pointer and 1 cycle cost since you can't use ldm on arm
23:11:47stripwaxneeds > 1 cycle to compute the pointer ..
23:12:20stripwax(unless you have bitreverse opcodes in silicon)
23:12:43saratogastripwax: j=revtab[k];
23:13:04 Quit ecaz ()
23:13:04saratogashould just be an index addressed load or whatever arm calls them
23:13:30saratogaoh but i guess thats really 2 loads
23:13:42saratogaone for the table and one for the actual data
23:14:16stripwaxis revtab a table of size 8192?
23:15:55saratogain the old fft its 2048 bytes
23:15:58saratogalet me see the new one
23:17:11stripwaxwow. tremor uses a 12-byte bitreverse lut, (so needs > 1 cycle to do it). a large bitrev table seems a huge waste of space, unless there's nothing else useful to put there
23:17:49stripwaxe.g. in iram
23:17:58saratogai think the new one is also 2048 bytes
23:18:08stripwax^12 byte^uh, 16 byte, obviously
23:18:39saratogatremor's bit reverse also uses almost 3MHz so i'm not sure its the best way to go
23:20:02stripwaxyeah, you can chop a chunk of that out by observing that you can do every-other bitrev12() call with just a single EOR. but doubt it makes a massive dent. tremors bitreverse also does butterflies inline and a bunch of other stuff, not just reordering the data.
23:23:39mtsaratoga: I hacked fft-fixed a bit to count muls and adds, and got some different numbers (4391 muls, and 10086 adds for fft512) .. I probably missed some add, I'll look for that, but muls count is close to yours, so it's probably correct.
23:24:05mtIn 2048 point fft though, I got > 23K muls (!)
23:24:18saratogamt: that sounds about right
23:24:26saratogawe only need up to 512 though
23:24:36saratogaat least for all current codecs
23:24:52saratogai think wma pro at 96kHz needs a 1024 point fft, but thats not a big deal for now
23:25:23mtOh ok, I was a little confused.
23:25:51stripwaxvorbis at 48kHz does too, iirc
23:26:06saratogai doubt 48k and 44.1 use different block sizes
23:26:07mc2739FlynDice: That patch seems to help - no more lockups on boot now.
23:26:12saratogathey're only a couple percent apart
23:26:58saratogamt: i *think* an N point transform return 2^N time domain samples, consumes 1/2*2^N frequency domain samples, and uses a 1/4*2^N point fft
23:27:24stripwaxsaratoga - 48k and 44.1 use different block sizes.
23:27:41saratogawhat does 48k use?
23:27:47stripwaxdon't forget that vorbis allocates the blocksizes at encoding time, not statically. but for <48kHz, the 8192 blocksize does not get selected
23:28:08stripwax48kHz can short+long blocksizes chosen from the whole range of blocksizes, selected at encoding time.
23:29:35saratogai realize that but IIRC all encoders always chose the same block sizes
23:29:46saratogasince 256 and 2048 are optimal at 48kHz
23:30:24stripwaxhm, encoders can't always choose the same blocksizes, or else we'd be able to remove the other window sizes from the decoder
23:30:53mtI have to go sleep. Goodnight.
23:31:09stripwaxthe reference encoder uses a 4096 blocksize for quality settings less than zero (e.g. q -1 ) apparently
23:32:12saratogamt: goodnight
23:32:28stripwaxsaratoga -
23:34:29saratogai wonder how well such a file decodes with rockbox now
23:45:52 Quit dfkt ("-= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.")
23:47:34saratogastripwax: my q-1 file decodes at 26.74MHz
23:47:47saratogawhich makes me think its not using a larger transform
23:48:11 Join funman [0] (n=fun@rockbox/developer/funman)
23:48:51saratogavorbis can only have 2 transform sizes per file right?
23:49:18 Quit p3tur (Remote closed the connection)
23:51:53funmansomeone with a c200v2 wanting to test new mkamsboot ?
23:54:54 Quit saratoga (Ping timeout: 180 seconds)

Previous day | Next day