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).

Notice: Only Gecko based browsers prior to FF4 support the multipart/mixed "server push" method used by this log reader to auto-update. Since you do not appear to use such a browser, this page will simply show the current log, and not automatically update.

#rockbox log for 2010-02-10

00:00:08bluebrotherpamaury: what are those "subtle properties" of FAT?
00:01:25pamaurybasically, when you do an opendir or more generally an open without modify any entry, one of the parameters is useless. This allows to get rid of the 30kib buffer used before. This is subtle because it's highly implementation specific (and I hope it's true actually ;))
00:01:55pamauryIf you want more details, I can go on...
00:02:20stripwaxSigh: run test_codec on a folder, but don't exit at the end. Then plug in USB, and delete the (0-byte length) test_codec_001.txt file in root. unmount and unplug USB: "Panic deleting zero length directory entry"
00:03:06***Saving seen data "./dancer.seen"
00:05:23 Join piotrekm [0] (
00:05:39 Quit captainkewllll (Quit: Page closed)
00:05:44 Quit piotrekm (Changing host)
00:05:45 Join piotrekm [0] (~pm@unaffiliated/piotrekm)
00:05:56 Quit piotrekm (Client Quit)
00:06:12 Join piotrekm [0] (
00:06:24 Quit piotrekm (Changing host)
00:06:24 Join piotrekm [0] (~pm@unaffiliated/piotrekm)
00:06:57bluebrotherpamaury: interesting. Though I guess it's too late for me to go into details :)
00:07:53 Join evilnick [0] (
00:08:45 Join akur [0] (
00:08:53pamauryInterestingly, this patch also augments the stack usage on my device when doing a full rebuild... I need to investigate that. Perhaps it's because I put fat_dir and fat_direntry on the stack at the beginning. If something as a device with lots of directory and (preferably) a very deep directory structure, I would like to see the result.
00:09:01 Join dmb_ [0] (~Dmb@unaffiliated/dmb)
00:12:58 Quit bertrik (Read error: Connection reset by peer)
00:13:19 Quit evilnick (Ping timeout: 256 seconds)
00:14:19 Quit Kitar|st ()
00:15:55 Quit pamaury (Quit: abort();)
00:21:25 Join Kitar|st [0] (
00:24:50flybackwell my 5 broken sansa's I got for $5 arrived
00:24:58flybackand I just realized I need sync cables to test anything
00:26:57flybackalso guys I got an error and if anyone has smart code in devel no matter how buggy, i'd like to try it
00:27:03flybackcause I fear the drop damaged the hd
00:31:06flybacki'll get the error in a bit
00:31:41flybackI bought the sansa's cause I heard rockbox has a usb debug driver so you can see what's on the lcd off the usb thru some kind of terminal
00:31:57flybackalso figured I could cut the dead lcd and possibly wire up another to the i/o pins
00:32:09CIA-88New commit by 03stripwax (r24575): Whoops, get rid of V, there is no V
00:33:03 Join evilnick [0] (
00:33:14flybacki'll seriously consider setting up a old p1, p2 linux box, with ssh login and sansa's left on it if any of you devels want somerthing to bang code off of
00:33:37stripwaxflyback - hm, didn't we talk about this the other day? there's no 'usb debug lcd terminal'.
00:34:31flybackwell I mean like message passing over usb
00:34:33 Join Tomis [0] (~Tomis@
00:34:36stripwaxthere's a usb serial debug but I think that needs to be enabled specifically (plus device needs to be running rockbox etc already)
00:34:49flybackno I am not talking about serial pins
00:34:53flybackI was talking about what you just said
00:35:06flybackbut thx that gives me an idea
00:35:13flybackI could probably jtag these also
00:35:15 Join DerPapst1 [0] (
00:35:27stripwaxjtag probably only way, in fact :-p
00:35:53flybackwell I got a wiggler jtag that does +5 and +3.3 targets
00:36:06flybackplus I bought a breakout board for a max30002 bidirectional logic level shifter
00:36:09flybackgoes down to 1.5
00:36:19flybackso I can buffer the jtag outputs if the jtag pins are lower than +3.3
00:37:00 Quit DerPapst (Ping timeout: 264 seconds)
00:37:22flybackreally neat little chip
00:37:31flybackonly paid $10 for the breakout board with the chip already mounted
00:37:41flybackup to 20mhz at some voltages and up to 40mhz on others
00:37:51flybackwhich is fast enough for a slow speed wiggler jtag
00:38:25flybackit's dual vcc so you don't have to deal with /direction pins and host/target modifications you just supply each side with the vcc for that side's logic level
00:38:33*TheSeven just spotted that there is no "Rockbox 3.5" choice for "Reported Version" in flyspray!
00:40:35 Join evilnick_ [0] (
00:41:18linuxstbTheSeven: That's because it's bug-free ;)
00:42:12linuxstbTheSeven: Err, I can see "Rockbox 3.5" there...
00:42:13 Quit togetic (Ping timeout: 256 seconds)
00:42:40 Join evilnick__ [0] (
00:42:59linuxstbTheSeven: Hmm, it appears as an option in the advanced search, but not when you add a new task...
00:43:21 Quit evilnick (Ping timeout: 256 seconds)
00:45:03 Quit evilnick_ (Ping timeout: 256 seconds)
00:48:32 Join PaulJam__ [0] (
00:49:01 Quit evilnick__ (Ping timeout: 256 seconds)
00:55:05 Quit DerPapst1 (Quit: Leaving.)
00:58:32 Join togetic [0] (~togetic@unaffiliated/ibuffy)
01:00:36 Join anewuser [0] (anewuser@unaffiliated/anewuser)
01:01:45 Join phanboy4 [0] (
01:02:19 Quit ender` (Quit: Life is what happens to you when you had other plans. -- John Lennon)
01:03:45 Join ansuz [0] (
01:04:43 Join evilnick [0] (
01:07:55 Quit ansuz (Remote host closed the connection)
01:10:44 Quit jgarvey (Quit: Leaving)
01:14:31 Quit evilnick (Ping timeout: 256 seconds)
01:16:51 Join evilnick [0] (
01:18:39 Join Tomis2 [0] (~Tomis@
01:20:56 Quit Tomis (Ping timeout: 252 seconds)
01:20:56 Nick Tomis2 is now known as Tomis (~Tomis@
01:21:53 Quit evilnick (Ping timeout: 248 seconds)
01:25:30 Quit piotrekm (Quit: Leaving.)
01:26:46 Quit Tomis (Read error: Operation timed out)
01:27:56 Join Tomis [0] (~Tomis@
01:29:16flybackif I don't make it, it's been fun
01:29:31flybackroads are bad and they accidnetley rmoetely powered off a server at work so I gotta drive over to hit the button
01:30:13krazykit`thanks for being offtopic
01:32:00stripwaxDoes anyone have sample vorbis files that contain 4096-sample or 8192-sample blocksizes? i.e. files encoded as -q -1 (negative quality) and/or files encoded from 64kHz or higher source material?
01:32:38stripwaxThink that we should add these to the test files for vorbis, since 4096 & 8192 blocksizes are something of an outlier but invariable someone complains whenever we (.. I) break vorbis decoding..
01:34:17 Quit shaggy-h (Ping timeout: 240 seconds)
01:35:31flybackkrazykit`, i might not be back, the roads are getting bad already
01:35:46flybackand if you actually read what I said, I said it's been fun being in this channel
01:35:48flybackbut whatever
01:35:51Mode"#rockbox +o rasher" by ChanServ (ChanServ@services.)
01:35:52Mode"#rockbox +q *!~teac@*" by rasher (~rasher@rockbox/developer/rasher)
01:35:56Mode"#rockbox -o rasher" by rasher (~rasher@rockbox/developer/rasher)
01:37:18 Part flyback
01:39:57 Quit MethoS- (Remote host closed the connection)
01:40:01CIA-88New commit by 03stripwax (r24576): Complete the job done in previous revision - now MDCT has NO INIT of its own. Only remaining init now is the revtab in fft. Removed swathes of ...
01:41:16*linuxstb wonders how stripwax can complete a job done
01:41:39*stripwax never completes anything, really, ever ...
01:41:57linuxstbSo "Continue the job started..." ?
01:42:15stripwaxMore likely "Finish the job started, I think"
01:42:24 Quit Farthen (Ping timeout: 264 seconds)
01:42:39linuxstbYou just said you never complete anything ;)
01:42:43*linuxstb thinks it's getting late...
01:42:52stripwaxhence "I think"
01:43:00stripwaxanyway. goodnight to you also :)
01:43:11 Quit stripwax (Quit:
01:43:13 Join MethoS- [0] (~clemens@
01:43:37 Join Farthen [0] (
01:47:03 Quit PaulJam__ (Ping timeout: 260 seconds)
01:55:38 Join Zarggg [0] (
02:00:38 Quit dfkt (Quit: -= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.)
02:03:09***Saving seen data "./dancer.seen"
02:05:35 Quit komputes (Ping timeout: 245 seconds)
02:10:54Unhelpfulstripwax: i have 96KHz files but nothing redistributable :/
02:13:08 Join evilnick [0] (
02:17:59 Quit evilnick (Ping timeout: 256 seconds)
02:20:44 Join evilnick [0] (
02:22:25 Quit akur (Quit: Leaving.)
02:23:47 Quit bmbl (Quit: Bye!)
02:24:16 Quit JdGordon_ (Quit: Page closed)
02:24:40 Join gigatheme [0] (
02:25:07saratogashould be easy enough to transcode the FLAC sample file into vorbis q -1
02:25:21 Quit evilnick (Ping timeout: 256 seconds)
02:25:47gigathemecan anyone here help me with the %Vi viewport tag for the .sbs files?
02:27:17 Quit maraz (Ping timeout: 265 seconds)
02:27:44CIA-88New commit by 03mc2739 (r24577): Thai translation update ...
02:28:20gigatheme%Vi|5|5|230|210|-|-|-| basically what does each section refer to?
02:28:57Stephen__Vi|x|y|length|height|font|fgcolor|bgcolor| afaik
02:29:09 Join maraz [0] (
02:29:19gigathemeok, is that for the text area then?
02:29:34Stephen__the list position
02:29:44gigathemeso i define the text's area then the images can fill the rest?
02:29:55gigathemewith being overwritten
02:30:01gigathemeor covered :P
02:30:06Stephen__use the %V for the images etc just like the .wps
02:30:25Stephen__the %Vi is just to define the list/menu
02:30:48gigathemeif i dont fill in the |font|fg|bg| does it pull those from the .cfg
02:31:14Stephen__yep you can just leave them as |-|-|-| if you like
02:31:18Unhelpfulwhat are our correctness tests for mdct_exp? "audio files don't sound obviously wrong"?
02:31:26gigathemethanks for the help :)
02:33:06 Quit gigatheme (Quit: CGI:IRC)
02:38:28 Quit Stephen__ (Quit: Leaving)
02:40:00 Join evilnick [0] (
02:42:25 Part froggyman
02:46:06 Join evilnick_ [0] (
02:48:01 Quit evilnick (Ping timeout: 256 seconds)
02:48:22saratogaUnhelpful: yeah
02:48:38saratogai guess before we commit we should compute RMS errors
02:52:47 Join CaptainKewl [0] (
02:54:08 Quit evilnick_ (Ping timeout: 248 seconds)
02:54:36 Join froggyman [0] (
02:55:28 Join histumness [0] (
02:57:06 Quit histumness (Client Quit)
03:09:03 Join gigatheme [0] (
03:10:56 Join evilnick [0] (
03:11:34gigathemehow do i specify a font for use on the .sbs?
03:12:02gigathemenot for the file browser, but like for the time and stuff
03:13:30linuxstbgigatheme: Rockbox only has two fonts - the small built-in system font, and the user-selected font. You can specify which one to use in the viewport definition
03:14:08gigathemeis that the difference between 0 and 1 in the %Vi
03:14:33gigatheme%Vi|12|20|216|222<<<would be here>>>|-|-|
03:14:53gigathememeh, i messed that up %Vi|12|20|216|222|<<>>|-|-|
03:15:03 Join S_a_i_n_t [0] (S_a_i_n_t@
03:15:27gigathemeso 0 means small system, and 1 is user? or opposite
03:15:28 Quit evilnick (Ping timeout: 248 seconds)
03:15:53S_a_i_n_tgigatheme: 0 system, 1 user
03:16:06S_a_i_n_tTorne: ping?
03:17:16 Join evilnick [0] (
03:21:25gigathemewait, this 0 or 1, that would apply to the file browser correct?
03:24:10S_a_i_n_tgigatheme: yes
03:25:01gigathemewell i want to keep the file browser in the user font, but i also want the time deisplayed in the small system font
03:25:18gigathemethe time will be in the .sbs
03:25:23S_a_i_n_tyou'd need to do it with viewports then
03:25:34S_a_i_n_tor an .sbs
03:25:51gigathemeyes .sbs is what im writing (first time :P)
03:26:40S_a_i_n_twell, if there is "user" font in the .sbs, but you want the time to be in "system" font...the .sbs need to draw another viewport fot the time
03:27:52gigathemei want to put the menu and files in the user font, then i need the time, and probably volume dB, and battery precentage in the system font
03:28:48 Quit evilnick (Ping timeout: 248 seconds)
03:29:14gigathemethe stock .sbs does this, im looking at it right here, but i need to figure out how its done
03:29:50S_a_i_n_twell, if you look hard's right there. you're looking at Cabbie, yes?
03:30:04gigatheme%Vi|0|8|-|-|1|-|-| this comes from the ""
03:30:18gigathemeno not for the .wps at all
03:30:29gigathemejust for the browser stuff
03:30:54S_a_i_n_tyes, the first two " - "'s mean the viewport is fullscreen, the 1 means it's in "user" font
03:31:30gigathemeok, and thats right, the files and menus are in the user font, but there is also a text clock in the top right which is in system font
03:31:31S_a_i_n_tnope...don't listen to me...that's completely wrong :P
03:32:33S_a_i_n_twhat they've done there is set a "UI Viewport, that uses the user font
03:32:45S_a_i_n_tanything outside of that will use the system font
03:32:52gigathemeah ok
03:33:11gigathemethat makes sense
03:33:37 Quit saratoga (Ping timeout: 248 seconds)
03:33:45S_a_i_n_tUI Viewport lets you specify an area of the scree nto display the UI that it isn;t fullscreen for example
03:34:13gigatheme%Vi|1|16|238|278|0| thats mine
03:34:48gigathemeso everything in the 238x278 square will be user font
03:34:54S_a_i_n_tso you'd need to look at the viewports that draw the clock etc, and change the 0 to 1, and it should all be in User font
03:35:26S_a_i_n_t*as an example
03:36:56gigatheme%?cc<%?ca<%?St|time format|<%cH|%cI>:%cM|−−:−−>|> thats from and it does what i want it to, but is there any reference to font there?
03:38:43S_a_i_n_tthe reference to font will be in the "main" viewport its dran in
03:38:59S_a_i_n_t*I think*
03:39:27gigathemewe'll i'll mess with it and see what happens
03:39:48gigathememaybe if i figure it out i can work on a new guide for the site
03:40:24S_a_i_n_tIt is a little outdated, well not really, it assumes a certain amount of knowledge is attained beforehand
03:40:34S_a_i_n_tthere isn;t really a "beginners guide"
03:40:44S_a_i_n_tbut there's an attempt at one :D
03:41:02gigathemeyeah, i thought maybe id be a little ahead since ive already written 3 .wps
03:41:12gigathemelol apparently not though
03:41:44S_a_i_n_teverything keeps changing with the .sbs's, that's *maybe* the reason it's been left alone for some time now.
03:42:01S_a_i_n_twhat's your target device btw?
03:42:09gigathemegigabeat f20
03:44:30S_a_i_n_tgigatheme: try messing with animation....that gets tricky :P
03:44:30 Join komputes [0] (~komputes@ubuntu/member/komputes)
03:44:43S_a_i_n_tI've got like 5/5 abandoned WPS's....LOL
03:44:57S_a_i_n_t5/6 rather
03:45:12gigatheme.gif animation
03:45:15S_a_i_n_tno, bitmap strips and conditional
03:45:25S_a_i_n_tbut I don't wanna put ideas in your haed :P
03:45:47gigathemesecond over is my latest
03:45:59S_a_i_n_tif .gif was'd take SOOOO much bloat out of my code
03:46:14 Join Zarggg_ [0] (
03:46:27gigathemewhat are you tring to animate? a clock?
03:46:55 Quit Zarggg (Read error: Connection reset by peer)
03:47:03*S_a_i_n_t grins
03:47:30S_a_i_n_tplaymode, AA transitional effects/ name it, it moves :D
03:48:01S_a_i_n_tyours is citrate?
03:49:23S_a_i_n_tyou've got some slight problems with the "magic" colour (255,0,255 magenta) on your shuffle and repeat icons...or was that intentional?
03:49:49gigathemeno, i accidently packaged the wrong file :P
03:50:13S_a_i_n_tjust reupload it with the same name, it'll replace the theme.
03:50:17 Join Rob2222 [0] (
03:50:25S_a_i_n_tyou'll get a confirmation email, then it''l get replaced
03:50:26gigathemewhat a novel idea! :D
03:50:43gigathemedidnt know that worked that way
03:50:50*S_a_i_n_t bows graciously
03:51:28S_a_i_n_tit was added *fairly* recently, I think because admins got sick of people asking "hay, I messe *this* up, can you replace X with Y"?
03:53:43 Quit Rob2223 (Ping timeout: 252 seconds)
03:53:47gigathemethat would get annoying
03:54:02 Join evilnick [0] (
03:55:56 Join evilnick_ [0] (
03:58:36 Quit perfectdrug_ (Read error: Operation timed out)
03:58:40 Quit evilnick (Ping timeout: 248 seconds)
03:59:27S_a_i_n_tgigatheme: I *finally* found the thread....missed it like 4 times looking for it.
03:59:32S_a_i_n_t should help
03:59:35 Quit phanboy4 (Read error: Connection reset by peer)
04:00:13 Quit panni_ (Read error: Connection reset by peer)
04:00:17 Quit evilnick_ (Ping timeout: 248 seconds)
04:00:21 Join panni_ [0] (
04:01:16 Quit panni_ (Read error: Connection reset by peer)
04:01:32 Join panni_ [0] (
04:02:03 Quit froggyman (Quit: Goodnight moon!)
04:03:10***Saving seen data "./dancer.seen"
04:06:10gigathemewhat does RTC mean :S
04:06:22 Join evilnick [0] (
04:06:37S_a_i_n_tReal Time Clock
04:06:56*S_a_i_n_t bows graciously
04:07:00gigathemethat makes sense :)
04:07:35gigatheme# Clock on RTC able targets, and disk access %V|-43|0|31|8|0|-|-| # align on the right with room for 5 SYSFONT digits %?cc<%?ca<%?St|time format|<%cH|%cI>:%cM|−−:−−>|>
04:07:41 Quit Tomis (Quit: Tomis)
04:08:00gigathemethats from the stock .sbs, and there it sets a viewport and uses 0 for system font
04:08:25S_a_i_n_tthat .sbs *probably* isn't the bast to learn it uses negative values for the viewport declaration
04:08:37S_a_i_n_twhich can get confusion pretty quickly
04:08:44S_a_i_n_twell, I found it did anyway
04:09:00gigathemeyeah thats kinda weird but i its just screen width subtract value = +version
04:09:25gigatheme240 - 43 = 197
04:09:31gigathemefor gigabeat are wise with the ways of the force young padawan :D
04:10:36gigathemestarwars fan?
04:10:42 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
04:10:43 Quit evilnick (Read error: Connection reset by peer)
04:10:46S_a_i_n_tnot really....
04:11:07 Join evilnick [0] (
04:11:09 Join saratoga [0] (~9803c6dd@gateway/web/freenode/x-lwemduzemodcajhr)
04:11:50S_a_i_n_tgigatheme: did you notice the #
04:12:04S_a_i_n_ton those lines the # means they aren't read you did.
04:12:21gigathemeyeah... i kinda got that already thanls though lol
04:12:59S_a_i_n_tsome people are like.."the code is right there!" why doesnt it work"...."remove the #"...."oh, right, lol"
04:14:42gigathemeif i set the clock viewport "%V||0|31|8|0|-|-|" with a foreground color, will it change the system font color inside of the clock viewport?
04:16:45gigathemeyou know off the top of your head (or deep within) how many pixels high the system font is?
04:16:46S_a_i_n_tI should've realised you've got a large DPI'd player when you were wanting to use the sysfont
04:17:11*S_a_i_n_t has a nano (well 5 of them actually), ans sysfont is almost IMPOSSIBLE to read :D
04:17:30S_a_i_n_t8 high, 12 wide
04:17:45S_a_i_n_t*I think*
04:17:53gigathemeseems high
04:17:53 Quit evilnick (Ping timeout: 248 seconds)
04:18:15gigathemeah, i cna measure a screendump
04:18:58S_a_i_n_tI think sysfont is just 08-Schumacher-Clean
04:19:04S_a_i_n_tpretty sure it is anyway
04:20:24S_a_i_n_tthere is *probably* a blank pixel on top/bottom of the font as well, to accomodate for "y" and such
04:20:39S_a_i_n_tbut I'm pretty sure sysfont needs an 8px line
04:21:07S_a_i_n_tand that each char is fixed width to 12px
04:21:42Unhelpfulsaratoga: don't at least vorbis and mp3 have mdct error-tolerance specs?
04:21:49 Quit TheSeven (Disconnected by services)
04:22:02 Join The_Seven [0] (~theseven@rockbox/developer/TheSeven)
04:22:03gigathemewell my top bar is 10px, so do i put it down 1y to center on the line, or do the invisible spaces make it 10 high and i put it a 0y
04:22:05saratogaUnhelpful: i don't think vorbis has any accuracy spec
04:22:11saratogabut mp3 does have conformance tests
04:22:12 Nick The_Seven is now known as TheSeven (~theseven@rockbox/developer/TheSeven)
04:22:23saratogabut they test the entire decoder
04:23:08saratogafor our purposes its probably sufficient to just test a WMA file with the new and old transform and a Vorbis file with 4096 blocks
04:23:22saratogabetween them, all block sizes will be used
04:23:28S_a_i_n_tgigatheme: 0y
04:24:00Unhelpfuli'd say rather than tests on the audio a set of test coefficients and expected transform outputs would be good. the "expected" results could be computed via mdct on doubles, and compared to the mdct_exp outputs... +/- 1 we can dismiss as rounding error. or maybe i'm making it more complicated than it needs to be :/
04:24:02 Quit panni_ (Read error: Connection reset by peer)
04:24:37saratogathe current transform is abotu as accurate as IEEE754 single precision, so it should be the same thing i think
04:26:02 Join evilnick [0] (
04:26:04Unhelpfulsingle is still 24 bits of mantissa, i believe?i was just thinking the "desired" result set ought to be generated with doubles to guarantee that no rounding errors (except from conversion to fixed-point of the output) are visible
04:27:57Unhelpfulremember, i optimize for the sake of optimization... and do the same thing w/ accuracy ;)
04:28:01saratogadoesn't hurt, but i don't think it matters
04:28:13saratogaas long as you're above ~10 bits no one can tell the difference
04:29:05saratogaI'd be pretty surprised if this version is less then 20 bits, baring algorithmic errors
04:29:51Unhelpfuland those ought to be glaringly obvious if we have compare to even outputs of our current transform
04:30:17saratogagiven the 1/2*log2(N) rounding error accumulation rule of thumb, we should expect ~26 bits
04:30:39saratoga(for fourier related transforms)
04:30:41 Quit evilnick (Ping timeout: 248 seconds)
04:31:07*Unhelpful is actually not familiar w/ this particular rule of thumb...
04:31:34UnhelpfulN = operations?
04:31:55saratogai've seen it around, but basically it just says you lose on average half a bit per level of divide and conquer in an fft like transform
04:32:01saratogawhich i guess is kind of obvious
04:32:55Unhelpfulah, it's transform-specific? i've only worked on one transform, and i only reimplemented ijg's C exactly.
04:32:56saratogai tend to think a split radix transform can do a bit better since its doing a lot fewer multiplies, but i'm not sure
04:33:31 Quit gigatheme (Quit: CGI:IRC)
04:34:34saratogaah yaeh, its for any sort of efficient fourier related transform
04:34:37Unhelpfuland i still wonder if perhaps a more mac-friendly algorithm might be possible... i had very few opportunities to use mac or armv6's packed dual-multiply-and-add
04:35:48saratogaisn't the packed MAC half accuracy?
04:37:33saratogaon armv6 i tend to think the biggest problem is going to be cache performance, since we don't even pay attention to that on PP/CF
04:37:49S_a_i_n_tiPod Nano 1g (not sure if it makes a difference), where in the Dubug Info should I be looking to see if the player knows whether or not it is charging?
04:38:27saratogadoes the debug screen actually show that?
04:39:04saratogai expect it to know if you're on external power, but i'm not sure it would know what the battery is doing since IIRC that target has a hardware charger
04:39:09S_a_i_n_tAccording to Torne
04:39:13 Join Tomis [0] (~Tomis@
04:39:23saratogawell if you know its there just look
04:39:55Unhelpfulsaratoga: all of the constants are 16-bit. were in the original code. as are the inputs.
04:40:06S_a_i_n_t%?bp and %?bc arent working...and he told me to check the debug screen to see if it was registering the charge there or not...
04:40:13S_a_i_n_tif it is there, I keep missing it.
04:40:13saratogafor MDCT or something else?
04:40:35 Quit anewuser (Quit: Another edition of chiptune gig WinterChip5! :O =ooo)
04:40:41Unhelpfulah, no, the transform i worked on for arm is ijg's jpeg idct
04:41:16Unhelpfulfor audio i'd say we can pretty much never use those lovely armv5e and armv6 bits :/
04:41:58saratogayou probably could, but it would require some effort
04:42:14saratogaa lot of the old ATRAC players for instance used 16 bit precision DSPs
04:42:22 Quit Sajber^ (Read error: Connection reset by peer)
04:42:41saratogaand very carefully chose where to use 48 bit software emulated operations and where to use 16 bit ones in the MDCT
04:44:12saratogaor can you not do a packed 16x16=32,16x16=32 mul?
04:45:47 Join evilnick [0] (
04:46:54Unhelpfulsaratoga: no, they're 16x16=32. there's also 16x32=48>>16
04:47:10 Quit Barahir (Ping timeout: 246 seconds)
04:47:27saratogayou could make that work in the mdct i think with pretty good accuracy
04:47:28Unhelpfulthey're all great if you can use them, since from arm9e on there's no early termination and the multiplier is 32x16
04:48:20saratogai thought 9E had a fully pipelined mul?
04:48:55saratogayou could rescale the trig constants so they're all 15-16 bit, do the pre/post mul in steps with different scale factors to account for the prescale
04:48:59 Join Barahir [0] (
04:49:01Unhelpfulsome multiply instructions still have worse than 1c throughput
04:49:39 Quit evilnick (Client Quit)
04:50:26 Quit n17ikh (Ping timeout: 256 seconds)
04:50:42Unhelpfulmull/mlal are 3c, mul/mla are 2c... i believe all multiplies with an operand that is explicitly half-word are 1c throughput
04:52:02saratogayeah but its fully pipelined, so its just an interlock and not a stall
04:52:02Unhelpfulhrm, except for smlalxy, but that's 16x16=32 with a 64-bit accumulate
04:52:41 Quit komputes (Ping timeout: 252 seconds)
04:53:13Unhelpfuli wouldn't think so, those multiplies all have a further cycle of latency, so i take the table to mean that it actually takes 2 cycles to do mul/mla
04:54:31Unhelpfularm11 gets even worse w/ latency, most multiply instructions have no results available for 2c after execution
04:55:07 Join n17ikh [0] (
04:55:23 Join midgey [0] (~tjross@rockbox/developer/midgey)
04:55:50Unhelpfuland mull goes from 1c delay for rdhi on arm9e to 1c delay for rdlo, 2c for rdhi :/
04:56:37saratoga"The ARM9E adds a 32 x 16 MAC unit (two-cycle latency with single-cycle throughput)"
04:56:54saratogaalthough i admit the pipeline diagram makes it look like there should be 2 cycle throughput
04:57:05Unhelpfulsaratoga: but 32x32 multiply (mul) is two operations on that unit
04:57:30saratogabut the manual says the first half happens in the execute stage, while the second half happens in the memory stage of the pipeline
04:57:36Unhelpfuli'm still tempted to try a goldschmidt divider on arm11 again, it wasn't quite as fast but i never really tuned it much... and it does much more in parallel w/ fewer stalls on results
04:57:48saratogaif it really used the multiplier twice i don't think they'd say the stage advanced
04:58:31Unhelpfulbut it's 32x16, how *can* it perform a 32x32 multiply in one pass?
04:59:10midgeysaratoga: i have some test_codec results for the mdctexp branch for the gigabeatf and e200 if you're interested
04:59:43saratogaUnhelpful: i assumed its actually two multipliers, one in the exec, and one in the mem stage
04:59:55saratogamidgey: sure, pastebin them
05:00:16 Quit MethoS- (Remote host closed the connection)
05:00:38Unhelpfulsaratoga: ah. i must admit we're at the limit of my knowledge of processor design here... the cycle timings i have came from the arm system developer's guide, which has them in very handy tables :)
05:01:02saratogayeah i looked at the tables now, just am not sure what they mean :)
05:01:10 Join vegtoruci [0] (
05:02:14Unhelpfulah, the tables in the ASDG are not very hard to follow, there's a "cycles" column that they pretty clearly mean to indicate throughput, and a "notes" column with any other caveats that mentions latency on the arm9e, and also early registers on arm11
05:02:53Unhelpfullonger multiply latency on arm11 *and* multiplier inputs as early regs make algorithms that do many chained multiplies rather bad on it :/
05:02:58midgeysaratoga: <- gigaf mdctexp
05:03:09midgey <- gigaf trunk
05:03:45midgeyi used r24576 for both, which is a bit unfair seeing as how trunk has had some optimizations to various codecs
05:04:58saratogamidgey: FWIW only vorbis and wma have the new mdct enabled
05:05:02saratogathe other codecs still use the old
05:05:31midgeywell then you have a lot of useless data :)
05:05:35saratogaboth vorbis and wma are 1-3 MHz faster
05:05:43 Quit FOAD (Ping timeout: 246 seconds)
05:05:50saratogamidgey: update the codec performance wiki page, no one has done the gigabeat f since 2007 IIRC
05:06:06midgeywill do
05:06:09saratogafor bonus points do the % improvement column
05:06:22saratogaUnhelpful: i guess you're right, thats really a let down
05:06:22Unhelpfulthe arm11 manual is similar, it says mul/mla are 2c, mull/mlal 3c, and the tiny muls are 1c. oddly the instructions that perform two 16x16=32 multiplies are *also* 1c... perhaps the multiplier on arm11 is "really" two 16x16 multipliers which may operate in parallel?
05:07:01saratogayes I assumed the option to do 2x 16x16 meant there was actually a pair of pipelined 16 bit multipliers
05:07:11saratogabut apparently not
05:07:24saratogai suppose this would explain why PP can keep up with the newer ARM targets
05:07:37saratogai always wondered why it doesn't get destroyed if the multiplier is really 4x faster
05:07:53saratogawell probably more like 2-3x faster if early termination is taken into account
05:08:11Unhelpfulsaratoga: it's still never *slower* than arm7tdmi, and you can have 1c multiplies if you can wait for the results and both inputs are 16bit, or if you want to multiply-and-shift a 32-bit value by a 16-bit one.
05:08:43Unhelpfulbut i suspect a *large* number of cycles in the new divider are spent stalled on arm11
05:09:06saratogai wonder why cook_stereo_32 is more then two times faster then cook_stereo_64
05:09:06midgeysaratoga: <- e200 mdctexp
05:09:14saratogai wonder if the files are misnamed
05:09:21midgeywith only wma and vorbis this time
05:09:42saratoga300.94% realtime for 128k WMA
05:09:55saratogamy first test of the codec was only 80% realtime on PP
05:10:12midgeyexcellent work all around
05:10:25midgeyi can test on h300 if you'd like, but i'd need someone to provide me with a build
05:10:40midgeyi haven't figured out how to compile gcc 3.4.6 on snow leopard yet....
05:11:24Unhelpfulok, bed for me :P
05:11:35saratogamidgey: lets wait until the XNPROD31_R ASM for CF is fixed
05:11:45saratogathe results will be a little slower right now
05:12:21midgeyyep, i assumed as much. just ping me or leave a message in the logs
05:12:21saratogai'll try and fix it tomorrow if stripwax doesn't first, i have to get some work done now
05:12:26saratogaok great
05:12:36saratogagood to know we're faster on the F
05:12:37 Join FOAD [0] (
05:12:48saratogalittle more work on CF and we'll probably be faster on all targets
05:13:02saratogathen i guess we check accuracy and roll out the lib
05:13:09midgeyi have a mini2g and ipod4g i can test on if it's useful
05:13:10saratogaand of course switch all the other codecs over to it
05:13:15midgeythose are both pp5022 iirc
05:13:22saratogaisn't the 4G 5020?
05:13:24midgeyso not too different than e200
05:13:27saratogathat might be neat, IIRC it has less IRAM
05:13:41saratogathough i guess still enough that the lib will perform the same
05:14:36midgeyyou're right on the 4g, it has a pp5020
05:19:47saratogamight be worth checking if it compiles and links
05:19:54CIA-88New commit by 03jdgordon (r24578): Fix FS #10983 - r24568 was just moronic. Sorry
05:21:13midgeysaratoga: i'm charging the 4g right now. i'll do a build with the same revision tomorrow
05:21:34saratogamidgey: just checked, it links fine, so probably not worth testing
05:22:20saratogathe only other really interesting targets to test on are the IPod < 4G and the gigabeat S
05:23:11midgeyunfortunately i don't have any of those, the pp5002 targets will be interesting
05:23:48saratogai'll do 3G later, need to get to work
05:23:56saratogaunhelpful can probably do the beast
05:25:17 Quit saratoga (Quit: Page closed)
05:41:32 Quit fyrestorm (Quit: lamers envy me like they envy bill g -- main boot xp, just the way it should be!)
05:44:38 Join flyback [0] (
05:53:08CIA-88New commit by 03jdgordon (r24579): OK, this is hopefully the last sbs related fix. This one will fix the backdrop going garbage, and add a missing else which kugel spotted
06:03:13***Saving seen data "./dancer.seen"
06:25:45 Join Guest23293 [0] (
06:25:55 Quit n17ikh (Disconnected by services)
06:25:59 Quit Guest23293 (Client Quit)
06:26:19 Join n17ikh [0] (
06:34:00 Quit dmb_ (Quit: Leaving)
06:35:50 Quit jd (Read error: Connection reset by peer)
06:36:08 Join jd [0] (
06:36:08 Quit jd (Changing host)
06:36:08 Join jd [0] (~jd@Wikipedia/HellDragon)
06:58:46 Quit midgey (Quit: midgey)
07:00:12 Join DomasoFan [0] (
07:02:17DomasoFanHi. does someone know if the Sansa Fuze with the following model is a Sansa Fuze v1?: SDMX14R-008GK-E57
07:05:02 Quit DomasoFan (Client Quit)
07:22:10 Quit HBK (Read error: Connection reset by peer)
07:22:43 Join grndslm [0] (
07:23:11 Quit CaptainKewl (Quit: ( :: NoNameScript 4.22 :: ))
07:23:15 Join HBK [0] (
07:49:30JdGordon multifont?
07:49:33JdGordonnot quite :p
07:50:06 Quit gevaerts (Disconnected by services)
07:50:18 Join gevaerts [0] (~fg@rockbox/developer/gevaerts)
07:57:21 Join ender` [0] (
08:03:15***Saving seen data "./dancer.seen"
08:10:12 Quit vegtoruci (Ping timeout: 260 seconds)
08:23:31 Join dmb [0] (~Dmb@unaffiliated/dmb)
08:30:39JdGordon <- working multifont
08:31:37 Join einhirn [0] (
08:31:54 Join shaggy-h [0] (
08:46:02 Join JdGordon1 [0] (
08:46:09 Quit JdGordon (Ping timeout: 256 seconds)
08:51:45 Join flydutch [0] (
09:00:47 Join petur [0] (~petur@rockbox/developer/petur)
09:01:09 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
09:06:39 Quit BHSPitMonkey (Remote host closed the connection)
09:12:40JdGordondoes anyone tihnk 10000 is too small for multifont for the fonts loaded by skins? (thats 10K each font)
09:13:03 Join kugel [0] (kugel@rockbox/developer/kugel)
09:15:02kugelJdGordon: I noticed you're changing the settings string directly
09:15:12JdGordonwhich? where?
09:15:26kugel*ptr2 = '|';
09:16:00JdGordonno, its copied to a buffer first
09:16:33kugelptr2 = global_settings.ui_vp_config;
09:17:17JdGordonyes, keep reading
09:17:30JdGordonlen = snprintf(ptr, remaining, "%%ax%%Vi|%s|\n", ptr2);
09:17:36JdGordonwhile ((ptr2 = strchr(ptr, ',')))
09:17:47JdGordonand ptr = buf; at the top
09:17:56 Quit Adub- (Quit: <Tobin> japanese men make hot trannies)
09:18:09kugelptr is the other buffer yes, but you're writing to ptr2
09:18:41 Quit JdGordon1 (Ping timeout: 248 seconds)
09:18:51kugelyou make the ui viewport setting look like "ui viewport: 5|1|30|2|2|ffffff|000000"
09:19:46pixelmado we really need an ui viewport setting anymore?
09:19:48JdGordonno, its ptr2 is pointing inside the buf buffer
09:20:42JdGordonpixelma: IMO we never did. but yeah, we may as well keep it
09:20:51pixelmawhich is a hidden setting and therefore only known by 10 people or so :/
09:20:52JdGordonare we looking at the same code?
09:21:23kugelptr2 is pointing to the settings string
09:23:08JdGordonwhich ois then snpriontf()'ed into ptr which points to the new buffer
09:23:30kugelthat doesn't change where it's pointing to
09:23:40kugelit's still pointing to the settings string
09:24:11JdGordonwhile ((ptr2 = strchr(ptr, ',')))
09:24:45JdGordonare you actually seeing the |'s in the .cfg file?
09:24:57kugelah right, I'm sorry
09:25:52 Join dmb_ [0] (~Dmb@unaffiliated/dmb)
09:26:29pixelmaI could understand dropping the ui viewport setting much much more than dropping any other theme menu items I can currently think of because colours or backdrop are easily set on device and the menu items speak for themselves. The ui viewport setting is hidden and if you would want to adjust it on device it's as complicated as editing an sbs file with just the %Vi
09:26:33kugeldid that commit also fix that the ui viewport is ignored?
09:27:17kugelpixelma: the ui viewport is for when you don't use a sbs
09:27:54pixelmabut it can easily replaced by using the most simple sbs
09:27:56kugelremoving it doesn't gain anything, it's impact on the code is about 10 lines
09:28:07JdGordon working multifont :)
09:28:37JdGordonkugel: the missing else? yes
09:28:50JdGordonpixelma: I want to add a UI for setting the ui viewport
09:29:13kugelJdGordon: I rather thought about the missing trailing | that as fixed
09:29:26JdGordonthats what caused the background garbage
09:29:33JdGordonI didnt realise the setting didnt have it
09:29:41JdGordonhence my commit message
09:29:45kugelwe had multifont "working" long ago so your screenshot doesn't thrill me
09:30:15kugelI haven't looked at the patch but after what you're saying I'm not a fan as well (but I won't object, I want multifont as well)
09:31:10pixelmanot sure if the intersection thing is still there which complicates things. And it also complicates understanding theme settings as you have two ways to achieve the same thing and might hit a situation where "a" doesn't work as you expect because "b" is in place and takes priority
09:31:41JdGordonthe intersection thing is gone
09:31:52JdGordonand /me pokes his tounge out at kugel :p
09:33:21JdGordonI dont know how you can not like this method. 0 wasted RAM if you dont use it (well almost 0 anyway), theoretically unlimited fonts
09:33:28JdGordonmakes the font cache more clever
09:36:49kugelI single buffer would be most clever
09:37:55JdGordonwtf? does the .glyphcache file only cache the glyph numbers? and not the actual glyph bits?
09:38:09JdGordon1 buffer would be MUCH more complicated
09:38:29JdGordonand not ever big enough to satisfy everyone, and too big for others
09:39:02kugelit could also grow and shrink with the number of fonts
09:39:13kugelthe lru mechanism works best in a single buffer
09:40:08kugelthe glyph cache probably only saves the index; it's just for letting the lru survive a shutdown
09:41:33JdGordonyeah, it makes sense, its just thats not what I ever thought it was doing
09:42:23JdGordonand no, a single buffer doesnt make lru any better
09:42:44JdGordonit might make the overall buffer more efficient though\
09:47:58 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
09:52:59 Join Bagder [0] (~daniel@rockbox/developer/bagder)
09:53:12 Quit kugel (Ping timeout: 264 seconds)
09:57:21 Join Speedy1 [0] (
09:57:59 Part Speedy1
10:00:15 Quit amiconn (Disconnected by services)
10:00:17 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
10:00:39 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
10:03:18***Saving seen data "./dancer.seen"
10:12:04 Join swilde [0] (
10:13:05 Quit AndyIL (Ping timeout: 248 seconds)
10:14:35S_a_i_n_tpixelma: I see your point with the UI Viewpoint, but I think not dropping it as an option is more to do with not forcing people to use an .sbs
10:14:48S_a_i_n_t*note the "I think* however.
10:15:20 Join Curulan [0] (~zarggg@2001:0:4137:9e74:0:96f8:beb1:ba3d)
10:15:26 Quit Zarggg_ (Read error: Connection reset by peer)
10:19:02 Join AndyI [0] (~pasha_int@
10:23:13 Quit AndyI (Ping timeout: 248 seconds)
10:23:52 Join AndyI [0] (~pasha_int@
10:33:42 Join PaulJam__ [0] (
10:38:28 Join Zarggg_ [0] (
10:41:24 Quit Curulan (Ping timeout: 260 seconds)
10:42:49pixelmaS_a_i_n_t: well, it's been discussed to drop other theme settings too. To be exact: not the setting but the menu items, the settings will still be there but hidden which is worse and the ui viewport setting didn't ever make it into the menu.
10:49:21S_a_i_n_tto be honest, I'm not much of a fan of dropping the other theme settings from the menu...
10:49:34S_a_i_n_tbut I do think UI Viewport should stay as a .cfg option
10:49:45S_a_i_n_tit's not doing any harm to those that don;t use it.
10:50:10S_a_i_n_tmany I'm having bad luck with ; vs. ' today...
10:53:02 Join _zic [0] (
10:53:41S_a_i_n_tOh, me saying "not dropping it as an option", I didn't mean a menu option (as I know there isn't one), what I meant was the option to configure it via .cfg that I think should stay.
11:06:36pixelmaI also want the other menu items to stay, at least as long as there is no other way of easily adjusting colours or stuff like that (via a plugin or so). But for ui viewport and an sbs with %we and your %Vi - I don't see a difference especially if the former stays a hidden setting - and then having two things doing the same can't be good.
11:13:15S_a_i_n_tthere's probably more than one instance of "more than one way to do the same thing" fact there is, by that respect, anything that has a .cfg setting should be booted out of the menu. Or perhaps sleep deprivation has my thinking somewhat backward as to what you're suggesting.
11:14:09gevaertsUhm, "anything that has a .cfg setting should be booted out of the menu" means "remove the settings menu"
11:15:28S_a_i_n_tthat was kinda my point, but a hint of sarcasm need be implied also.
11:15:40S_a_i_n_tmore than a hint actually
11:17:14S_a_i_n_tlong story short, IMHO UI Viewport should stay in the .cfg until there's a better way to do it, and probably even if there is.
11:22:19pixelmahuh? I think that's reverse - your "even" would be my only "if that happens then there might be one reason for it to stay". SBS is already another way to do it, not better but not worse (which is my point that other settings being cfg one only will be worse than what there is now, try chosing a different background colour if it is only a cfg file setting, getting your hex number out of thin air and only seeing the result after getting back and forth
11:22:19pixelma a few times usig the text editor.
11:22:45AlexPI did a battery bench on my clip last night - 09:02
11:24:46pixelmano-one explained to me yet where the difference between ui viewport and such a simple sbs is (other than maybe the %we line), adjusting coordinates on target would be exactly the same
11:25:05S_a_i_n_twhat I meant by "even if there is" is that it probably has the potential to give a lot of support related grief, like "why is my menu now only 1px wide" so on and so forth.
11:25:36S_a_i_n_tIt's probaby better (for now) if themers handle the UI Viewport
11:26:25S_a_i_n_tand the difference is, not forcing people to use .sbs
11:27:01S_a_i_n_tat the moment, if you want UI viewport, but not an can.
11:27:51pixelmabut the sbs would only be "use this %Vi" as ui viewport, nothing else
11:27:51S_a_i_n_tdon;t get me wrong, I can see your point (and I think it's valid too), but there are two sides to a coin.
11:28:21S_a_i_n_tbut it still forces an action, which isn't good IMO
11:28:58*gevaerts still thinks that *if* independently designed .sbs and menu backdrop are supposed to be supported, the intersection method is the only working solution
11:29:01pixelmawhat action (more than setting an ui viewport)?
11:30:51S_a_i_n_tthis could go back and forth for a long time, so /me decides to drop it.
11:31:10pixelmaI mean compared to editing the cfg file - where is a difference in "action" to editing an sbs file?
11:31:17S_a_i_n_twhilst retaining his opinion of course.
11:31:36S_a_i_n_thaving to have an .sbs in the first place.
11:31:45S_a_i_n_tEVERYONE has a cfg
11:31:55S_a_i_n_ta LOT of people don't have .sbs's
11:32:29S_a_i_n_t(apparently dropping it isn;t an opyion) :P
11:32:38pixelmayou would have to add a ui viewport line (and have to know the syntax)
11:33:03S_a_i_n_tUI viewport line is always in the .cfg
11:33:08S_a_i_n_tjust with no options
11:33:17S_a_i_n_tui viewport: -
11:36:10pixelmaonly if you save a full cfg and then I still have to know the syntax for the coordinates etc.
11:36:46S_a_i_n_tyou need to know the syntax for .sbs's totally double edged
11:36:54S_a_i_n_tnooone's going to be happy apparently
11:38:21S_a_i_n_tbe back in 5 mins...
11:38:23 Quit S_a_i_n_t (Quit: [St.] has exited mIRC™)
11:39:02pixelmayes, which is why I don't see a difference and sbs giving more power
11:39:58 Quit bug2000 (Read error: Operation timed out)
11:40:45gevaertsUnless the sbs specifies the backdrop to use (does it these days? I haven't really followed closely), there should be a not-in-sbs way to specify the viewport
11:42:02gevaertsThat way *should not* be used by themes, it's the way you adjust things if whatever viewport settings that come with the sbs/skin are not compatible with the backdrop the user selected
11:42:06pixelmait can
11:42:29 Join S_a_i_n_t [0] (S_a_i_n_t@
11:43:30pixelmabut if you don't specify a backdrop, the main backdrop set through the menu is used (or that was the plan)
11:43:34 Join evilnick [0] (~evilnick@rockbox/staff/evilnick)
11:44:11*S_a_i_n_t thinks he's missed some of this conversation...
11:45:35*pixelma points S_a_i_n_t to the topic
11:45:45liari dont hear sound from rockboy - is it disabled for my target? (nano2g)
11:45:54TorneS_a_i_n_t: i did tell you exactly where the battery status was btw :)
11:46:11TorneS_a_i_n_t: if you didn't find it yet, the secret is that the "view battery" screen, with the graph, has multiple pages: scroll the wheel to see the others
11:46:15S_a_i_n_tin debug isn;t that exact...
11:46:16gevaertsok. My view is that as long as that is supported, there should also be a non-sbs way to specify the viewport, and I think it should be as easy to use as the backdrop setting (i.e. not .cfg-only)
11:46:43S_a_i_n_tgevaerts: exactly
11:46:46TorneS_a_i_n_t: I told you to scroll down on the view battery screen
11:46:56S_a_i_n_tgevaerts argues a point a LOT better than I do apparently
11:46:58Torneso yes, it was exact :)
11:48:14gevaertsMy reason for that is that you want to be able to update a theme from the theme site without having to re-do half of your local settings again (the other half being the backdrop itself)
11:49:14S_a_i_n_tTorne; I couldn't tell what it was I was supposed to be looking at "battery, power status, voltage delta or Last PowerHist)
11:50:24Torneis it not self-explanatory?
11:50:38S_a_i_n_tapparently not.
11:51:11Tornethe screen i was directing you to is ifdef'ed for IPOD_NANO || IPOD_VIDEO
11:51:48S_a_i_n_tyeaH, someone else said your theory wouldn;t work (scorche maybe?) but I wanted to ask you...
11:52:02S_a_i_n_tF'N SEMICOLON!!!
11:52:03Tornethat is odd, though, i'm pretty sure it should be available
11:52:23Tornewhich model is yours again?
11:52:34S_a_i_n_tand I'm guessing *that's* what %bp and %bc don;t work on the Nano
11:52:42S_a_i_n_t1gen Nano
11:52:43TorneWait, you have a nano1g?
11:52:49S_a_i_n_thavent tried the 2geb yet
11:52:51Tornethe debug screen is enabled on IPOD_NANO
11:52:55S_a_i_n_tbut I assume it's the same
11:52:55Torneso it should be right there
11:53:04Tornei thought you had a mini or something
11:53:22S_a_i_n_t5 in fact
11:53:25TornePower status screen
11:53:42Torneshould list USB pwr, EXT pwr, Battery
11:53:54TorneUSB pwr should be obvious..
11:54:01Tornebattery says charging/charged/discharging
11:54:05Tornewhich should be obvious also :)
11:54:13Torneext pwr I believe is the 12V from the firewire pins?
11:54:53Torneare those lines not there for you?
11:55:31TorneWell, er
11:55:34Tornethey are supposed to be
11:55:45S_a_i_n_tnothing changes in the power status screen except the voltage jumps slightly
11:55:54S_a_i_n_tafter watching it for a few seconds though
11:55:59TorneThose lines should be there all the time
11:56:06S_a_i_n_tbut the lines your talking about aren;t ther at all
11:56:14Tornewell they're defined for nano and video
11:56:23Tornelikewise the lines in power-ipod.c which return the charger status
11:56:32Tornewhich should be what drives the wps tokens
11:56:42S_a_i_n_tI can screendump if you want me to confirm I'm not an may help
11:56:45Torneso yes i am pretty certain it should work.
11:56:46S_a_i_n_tjust in case am
11:57:15Tornethe screen starts with "Power status:"
11:57:24Tornethen a line with "Battery: 3.6V" or whatever
11:57:32Tornethose are unconditional for all targets so that *must* be there :)
11:57:48Tornethere should be like 5-6 more lines
11:57:48S_a_i_n_tyes, and it just says Battery: *voltage
11:58:28Torneno need..
11:58:32Tornei have a theory
11:59:17Tornehm. mayhbe not.
12:00:14Torneback in a sec
12:01:04 Join S_a_i_n_t_ [0] (S_a_i_n_t@
12:01:31S_a_i_n_t_shit, sorry...pressed X instead of minimise
12:02:12S_a_i_n_t_I have a file on the root of my player called identify_info
12:02:17S_a_i_n_t_not rb related?
12:02:27S_a_i_n_t_I've never seen it before
12:03:19***Saving seen data "./dancer.seen"
12:04:34*gevaerts accuses S_a_i_n_t_ of having selected random debug menu items :)
12:05:07S_a_i_n_t_ssssshhhhh you :P
12:05:56S_a_i_n_t_and no, it's not photochopped :P
12:08:34S_a_i_n_t_Hmmmm...ok, so identify_info *is* Rb related...but what is it?
12:08:40S_a_i_n_t_can I kill it?
12:09:12 Join kugel [0] (kugel@rockbox/developer/kugel)
12:16:27 Join archivator [0] (
12:18:58archivatorCould someone please take a look at fs#10065 and my last comment on there? I would really like to see that task closed..
12:21:03gevaertsarchivator: if nobody beats me to it, I'll have a look tonight
12:22:00archivatorThanks, gevaerts :)
12:22:11 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
12:22:56TorneS_a_i_n_t: OK, I cannot find any reason why the nano1g shouldn't display that info.
12:23:09TorneS_a_i_n_t: it is indeed not present on any *other* ipod doesn;t.
12:23:20S_a_i_n_t_guess I need to file a bug report then?
12:23:21Torneonly nano1g and video (nano2g has its own different info there)
12:23:52TorneThere we go :)
12:23:56Tornenano1g doesn't define CONFIG_CHARGING
12:23:59Tornewhich is why none of this works
12:23:59S_a_i_n_t_found it?
12:24:11*S_a_i_n_t_ was beginning to think he was going insane
12:24:38Tornethe 4g/color/mini1g/mini2g/video all have CHARGING_MONITOR
12:24:42Tornebut nano1g has nothing
12:25:05S_a_i_n_t_*should* it have it?
12:25:12S_a_i_n_t_I'm guessing so
12:25:25TorneI was under the impression that all those models, including nano1g, had approximately the same charging arrangement
12:25:37Tornesame PCF, same charge controller, just mayhbe different GPIO pins used
12:25:41Tornebut i might be wrong
12:26:13Tornethe code for it is there in power-ipod.c
12:26:14Tornebut it's ifdef'ed out as a result of not having CONFIG_CHARGING
12:26:32Tornetry setting it to CHARGING_MONITOR in the config
12:26:35Torneand see if that even compiles
12:27:07S_a_i_n_t_that may take a while...I'll get back to you :P
12:27:23Tornewell i can see if it compiles, if you want, but i can't test it :)
12:27:32 Quit liar (Ping timeout: 252 seconds)
12:27:57S_a_i_n_t_then it may as wel be me, I can do both
12:28:03S_a_i_n_t_it'll just take longer
12:28:32Tornemeh, i'm building it now
12:29:01S_a_i_n_t_send me the .diff if it works then :P
12:29:16Torneyou can just have the build
12:29:21Torneif you want to try it :)
12:29:50S_a_i_n_t_thank the "friendly neighbours"
12:30:18S_a_i_n_t_dammit, my tranfer rate is shocking was supposed to come before that
12:30:48 Join teru [0] (
12:31:29Tornelinuxstb: do you know why nano1g doesn't have CONFIG_CHARGING defined?
12:31:40Tornerockbox builds, i'll let it do the rest and make you a zip
12:32:28S_a_i_n_t_upload it somewhere, don't bother with'll get pissed off by my shitty transfer rate
12:33:07*S_a_i_n_t_ solved his semicolon problem by remapping the keyboard, prying both keys off and switching them :D
12:33:18S_a_i_n_t_better than learning to type properly :D
12:33:35gevaertsTorne: r16124 might half-explain this
12:33:39Torneupload? doesn't every computer everywhere have a webserver?
12:34:46S_a_i_n_t_this is a "....buh?" moment for me.
12:34:52Tornegevaerts: hm. the video/nano have different gpios
12:35:01Tornebut the code implies strongly that the video/nano work identically
12:35:12 Join piotrekm [0] (~piotrek@unaffiliated/piotrekm)
12:36:13gevaertsSo the question is why it wasn't enabled by r8966
12:36:19Torner19315 added nano support to power-ipod.c
12:36:23Tornebut didn't change CONFIG_CHARGING
12:38:09linuxstbTorne: I've no idea. Has it never been enabled?
12:38:15Tornelinuxstb: it's neve rbeen enabled
12:38:30Torner16124 implemented charging monitoring for 4g/mini/color
12:38:37Tornevideo came at some differnet point
12:38:48Tornethen 19315 changed the ifdefs to make the video code cover nano as well
12:38:58Tornebut nothing ever turned on CONFIG_CHARGING for nano so that code isn't used
12:39:06Tornei guess S_a_i_n_t can try it for us
12:39:18Tornewhat else does CONFIG_CHARGING affect? is there any chance it could make stuff go horribly wrong?
12:39:47S_a_i_n_t_bah....nano's (ipods in general) are unbrickable
12:40:02S_a_i_n_t_besides the hammer (and my newfound coffee) method
12:40:22S_a_i_n_t_that nano never recovered....R.I.P
12:40:49linuxstbTorne: I've no idea about the power management stuff. I wouldn't have expected it to make much difference on the ipods, apart from cosmetic things like charging indicators.
12:41:21TorneYah, it's only monitoring
12:41:28rasherThat's not insignificant
12:42:13S_a_i_n_t_so this *is* the reason my %bc/bp tags are broken?!?
12:42:20Torneyes, this is exactly the reason
12:42:35Tornecurrently the nano1g does not support *any* kind of charging status whatsoever
12:42:40Tornebecause it's not been turned on :)
12:42:45*S_a_i_n_t_ finally found a bug of interest, and not annoyance
12:42:54Tornethe code is there and it compiles
12:42:59Torneso i guess it's just whether it *actually works*
12:43:05Tornei.e. displays the right status
12:43:16S_a_i_n_t_I'll let you know in 10mins or so.
12:46:52Tornehave you got something that provides 12V firewire power, also?
12:47:07Torneif so it would be good to check whether the EXT power changes appropriately when you insert that
12:47:12Tornecar chargers usually do?
12:47:44S_a_i_n_tno, and nano isnt firewire anyway.
12:48:29TorneNo, but it still accepts 12V power on the pins used by the firewire cable
12:48:32Torneall the ipods do
12:48:42Tornethe data lines don't exist, no
12:49:01TorneFS #6891 suggests that the code *isn't* right and that the nano has different GPIOs for this
12:49:04 Quit Tomis (Read error: Operation timed out)
12:49:09Torneso yes there is alrady a bug for this
12:49:25 Join Tomis [0] (~Tomis@ this wont work?
12:49:45TorneIt might
12:49:46 Quit PaulJam__ (Read error: Connection reset by peer)
12:49:48TorneIt depends who is right :)
12:50:05S_a_i_n_t_we'll see very soon
12:50:10TorneFS #10591 has a patch which does exactly what i have just done, dreamlayers evidently was also assuming tha tthe code already in svn works
12:50:45TorneFS #6940 is also relevant
12:50:46Torneso, hm
12:50:53Tornethere have been three different patches that claim to fix this
12:50:54Torneall different :)
12:50:56TorneHOORAY :)
12:51:09S_a_i_n_t_yay! ....not
12:52:18Torneso i guess the next question is does anyone have a nano1g and both usb/firewire chargers and can get a multimeter appropriately attached? :)
12:54:24TorneOh wait, no
12:54:34Tornethe patches are all using the same GPIO pins as the current code after all
12:54:46Tornethe ipodvideo code got changed since these patches were written :)
12:55:23 Quit Hadaka (Ping timeout: 276 seconds)
12:55:56Torneso if this works for you we can probably commit it and close out all three of those bugs
12:58:12 Join watto [0] (~watto@
12:58:51S_a_i_n_t_fingers crossed...
12:59:35S_a_i_n_t_(usb) is slooooooow for some reason...
13:00:30Torneusb transfers being slow on ipods is a known issue, no?
13:00:43Tornethe rockbox build is made of loads of tiny files, which is a worst-case for USB mass storage anyway
13:01:11S_a_i_n_t_yeah...nano2g is about 100% faster than 1g
13:01:34S_a_i_n_t_for transfer that is.
13:01:52GaloisI never experienced much speed difference between the two
13:02:12linuxstbTorne: Apple's emergency disk mode is known to be slow on the nano1g and video.
13:02:14S_a_i_n_t_i have, DRAMATIC difference
13:02:36GaloisI don't use the emergency disk mode, but I do use the disk mode
13:02:45Galoisor are they the same
13:02:47Tornelinuxstb: wait, on the video?
13:03:02S_a_i_n_t_Galois: yep
13:03:03Tornei've not noticed it being slow
13:03:10Torneahwell *shrug*
13:03:23Tornei was talking about using rockbox usb, anyway
13:03:45Galoiswell, either way, rockbox usb or disk mode, it's as fast as any other usb device for me was i
13:04:02linuxstbGalois: "emergency disk mode" is the disk mode built into Apple bootloader in the NOR flash (accessed by pressing SELECT+PLAY when booting, or if Rockbox reboots into it). The other disk mode is the version in Apple's main firmware.
13:04:29*gevaerts ignores all USB speed reports that do not have actual numbers and a description of the testing method
13:04:41 Join Hadaka [0] (
13:05:52 Join liar [0] (
13:05:54S_a_i_n_t_Torne: sorry man, debug still has the one option
13:06:17S_a_i_n_t_battery that is... :D
13:06:20Tornethat's definately impossible ;)
13:06:37S_a_i_n_t_er...hang on
13:07:25S_a_i_n_t_you fixed it!!!
13:07:25 Join MethoS- [0] (~clemens@
13:07:43Torneyou mean the wps tokens work?
13:07:50Tornewhat about the debug menu?
13:08:02S_a_i_n_t_usb power, ext power, battery, charging, headphone!
13:08:09TorneYup, that's what's supposed ot be there :)
13:08:25S_a_i_n_t_my ipod did a "first something"
13:08:30*S_a_i_n_t_ feels speciasl
13:09:03S_a_i_n_t_wow....3 bugs with one commit...that's gotta be a first, or close to a record :D
13:09:12TorneWell, don't jump the gun, not that it's committed yet
13:09:22Tornei presume usb power toggles on and off when you plug and unplug the cable?
13:09:26Torneand battery says charging/discharging?
13:09:35S_a_i_n_t_it sure does
13:09:46Tornecan you check that it correctly reports when the battery is charged?
13:10:11Tornebest method to do this woul dbe boot the OF and let it charge there until the OF says charging is done, then start RB again and plug cable back in, give it a few mins
13:10:44S_a_i_n_t_ok, but it may take a while...
13:10:48S_a_i_n_t_will do however
13:10:50Tornei'm willing to assume the firewire power pin is handled right, since it's done the same on video and in the patches for the other issues, but none of them address the charged status
13:10:52 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
13:10:58Tornewhich is different code
13:11:56Torne(check both the debug menu state and the wps for that, btw)
13:13:10S_a_i_n_t_ok....wil do, disk-mode god enough for the "charg in OF" bit? or actually boot into the OF necessary?
13:13:16 Join DerPapst [0] (
13:13:24Tornedisk mode will do if it actually tells you when the battery is charged
13:13:26Tornei dunno :)
13:13:38TheSevenit does so on the 2g at least
13:14:03*S_a_i_n_t_ feels usefull for a change
13:14:12TheSevenis there really a "charged" status reported by the hardware? i have the impression that on the 2g they just determine that by the battery voltage and current readings
13:14:12*S_a_i_n_t_ likes this new feeling :D
13:15:03TorneTheSeven: well, the code for the PP ipods does GPIOB_INPUT_VAL & 0x01
13:15:07Tornefor all models apart from Color
13:15:16Torne(and 1g/2g)
13:15:26S_a_i_n_t_Hmmm....seperate topic, but last time I checked I noticed that the 2g reported "estimated time left" as 4000 minutes or so...that can't be right? no?
13:16:14TorneTheSeven: and as far as i've observed that reports accurately for the video
13:16:25S_a_i_n_t_4882 minutes even, at 100%
13:16:28Tornesee charging_state in power-ipod.c
13:16:30*TheSeven blames that on nobody having calibrated it yet
13:17:09*TheSeven would actually like to do a rework of the whole rockbox power management code
13:17:37Tornewell off you go :)
13:17:47S_a_i_n_t_I was waiting for that :P
13:18:03Tornei'm only going to rework usb charging
13:18:09Tornethe rest will have to be someone else ;)
13:18:28 Join Zagor [0] (~bjst@rockbox/developer/Zagor)
13:18:43TheSevenwe need to account for things like the inner resistance, and i would like to base the runtime estimates on real current readings for targets that have them
13:19:24Tornethe latter would be nice, but i'd be a bit cautious about how you average ti
13:19:49Tornee.g. playlists that include a variety of codecs with drastically different boost ratios
13:19:55TheSevenTorne: yes, one of the problems of that is that measuring the current will cause additional current to be drawn, and thus included in the measurement
13:20:03 Quit DerPapst (Ping timeout: 272 seconds)
13:20:09Tornewell, i didn't mean that :)
13:20:11Tornebut yes
13:20:28Tornei mean more the case where you have an album in something expensive, and one in something cheap..
13:20:38Torneis the estimate going to shift by many hours depending which album the playlist has reached? :)
13:20:40TheSevenTorne: codecs won't really matter as mp3, wma, vorbis and flac all play realtime without needing to boost
13:20:57TheSevenand i don't think other codecs are popular enough to be an issue...
13:20:58Tornealso the disk spinup frequency with flac..
13:21:01Torneand so on
13:21:04Tornei dunno. i'm just speculating
13:21:09Torneit may be not that much of a problem
13:21:48TheSevenanyways, I think it will always be more accurate than what we're currently doing, which is just assuming that it will always take the same current
13:22:14TheSeventhe bigger problem is, that when charging, the battery voltage will quickly be at 4.18V, but that doesn't tell *anything* about the charge it has taken with li-po batteries
13:22:38*S_a_i_n_t_ is *really* glad he took this problem to IRC instead of the forums... ;
13:23:20 Join DerPapst [0] (
13:23:30TheSevenwe could fix that by finding out the approximate inner resistance of that battery and "fixing" the voltage reading according to the current
13:24:00*TheSeven needs to check whether that is linear for lipos though
13:24:08S_a_i_n_t_and *really* glad for the existence of the lifeform known as Torne
13:24:09Tornewell, that part is definately beyond me :)
13:30:52 Part wookey_
13:37:26 Join perfectdrug_ [0] (
13:37:53 Quit antil33t (Read error: Connection reset by peer)
13:38:00 Join antil33t [0] (
13:39:04 Join Omlet [0] (
13:53:17 Quit Galois (Remote host closed the connection)
13:53:40 Join Galois [0] (
13:53:50 Quit Galois (Client Quit)
13:54:26Tornewow, someone uses frotz a lot, on the forums
13:54:36 Join Galois [0] (
13:55:50CIA-88New commit by 03mc2739 (r24580): Updated Portuguese Translation ...
13:56:00S_a_i_n_t_i do too, i just don't post about it. for me, IF on my ipod is better than...errr, somethingelse hat's good ;)
13:58:17Bagderdid anyone ever try ffmp3 with rockbox?
13:58:28Bagderthe ffmpeg guys says its faster than libmad at least
13:58:59TorneS_a_i_n_t_: actually playing existing IF wasn't what i had in mind for frotz at all, but hey, it's a freebie feature :)
13:59:29S_a_i_n_t_what did you have in mind?
13:59:33rasherBagder: So's our decoder :)
14:00:04Bagderyes, but our is libmad based
14:00:20kugelyes, and saratoga constantly complains about it
14:01:14 Join Sajber^ [0] (
14:01:54*kugel is scared by
14:03:20***Saving seen data "./dancer.seen"
14:08:49 Quit AndyI (Ping timeout: 248 seconds)
14:11:23 Join GeekShado_ [0] (
14:11:32*Bagder talked to the friendly people in #ffmpeg-devel
14:11:46BagderI'll post a follow-up on the mailing list on that
14:12:40 Quit linuxstb (Ping timeout: 245 seconds)
14:13:55 Quit GeekShadow (Ping timeout: 265 seconds)
14:14:06 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
14:16:05gevaertsTorne: so you feel all useful now? :)
14:17:20Tornegevaerts: heh, it's nice to know that somoene cares
14:17:33Torneand it might encourage me to implement more of the missing features sooner
14:17:46Torneand if jd gets multifont to work.. ;)
14:18:36TorneS_a_i_n_t_: I was intending to use it as a platform to actually develop menu/button based games
14:18:53TorneS_a_i_n_t_: e.g. things with JRPG-style combat
14:19:08Torneso, text based, but not text adventures in the usual sense
14:19:24Torneor, choose-your-own-adventure style things
14:19:34Tornethere are various libraries for inform that help to do some of that
14:19:36linuxstbkugel: Are you sure saratoga complains about our mp3 decoder? I thought he was happy with it, especially as he has made it dual-core. I know he (and anyone else who has looked at it) complains about libfaad (the AAC decoder), are you perhaps confusing them?
14:19:38Torneand i was going to maybe write some more.
14:20:09Tornethus it might eventually be possible for people with limited programming ability to write certain kinds of games for rockbox :)
14:21:48 Join AndyI [0] (~pasha_int@
14:22:43kugellinuxstb: yes, dual core makes it fast on PP, but not on the rest of the targets
14:23:08rasherI thought preglow took care of that long ago?
14:23:12 Quit Rob2222 (Quit: Rob2222)
14:23:59kugelmp3 is slower than wma and ogg on the rest
14:25:29 Quit AndyI (Read error: Connection reset by peer)
14:26:13Bagderso maybe ffmpeg's codec is an idea after all
14:27:55 Quit piotrekm (Quit: piotrekm)
14:28:06 Join AndyI [0] (~pasha_int@
14:30:06 Quit Tuplis (Quit: hetkinen...)
14:30:15S_a_i_n_t_Torne: what am I doing? OF reports the battery is charged, so I connect under RB and see if RB thinks it's charged (after some minutes) also?
14:33:08Torneit might sag slightly from 100% while you reboot to RB
14:33:15Torneso you might have to give it a sec to finish topping up the battery
14:33:23Tornebut it should hopefully report charged within a short time
14:33:26S_a_i_n_t_commit away
14:33:39TorneWell, that sounds good to me :)
14:33:50S_a_i_n_t_me too ;)
14:33:58Tornethe behaviour of the code in svn is already identical to the past patches for this issue
14:34:04Torneso the only change needed is to actually turn monitoring on.
14:34:16Tornei shall do it momentarily :)
14:34:42S_a_i_n_t_thanks made a feature of my wps...and cabbie actually, actually work!
14:35:07S_a_i_n_t_does this mean i gt a spot in the credits?
14:36:24S_a_i_n_t_that, (the monitoring thing) sems WAY too obvious a thing for the ammount of people that "apparently
14:36:44S_a_i_n_t_looked at it" to miss...o?
14:36:56Tornewell those FS# all had various people propose fixes
14:37:24Tornebut i would assume that the intersection between "people who could test it" and "people who are committers" and "people who noticed the FS# entries" was probably empty :)
14:37:48S_a_i_n_t_ahhhh...then came me :)
14:39:11 Join robin0800 [0] (
14:39:31S_a_i_n_t_I *still* (and probably forever will) find it *really* cool that this Nano was a first to do something :D
14:39:44*S_a_i_n_t_ pats his trusty 1g Nano affectionately
14:39:51CIA-88New commit by 03torne (r24581): Enable charging monitoring for the Nano 1G. ...
14:40:16S_a_i_n_t_and scours at the bits of his coffee-ridden 2g
14:45:05 Quit shaggy-h (Ping timeout: 240 seconds)
14:46:52 Quit linuxstb (Ping timeout: 260 seconds)
14:50:59 Join shaggy-h [0] (
15:09:35 Join dfkt [0] (dfkt@unaffiliated/dfkt)
15:15:04 Quit S_a_i_n_t (Disconnected by services)
15:15:57 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
15:15:59 Nick S_a_i_n_t_ is now known as S_a_i_n_t (S_a_i_n_t@
15:16:52 Quit HBK (Read error: Connection reset by peer)
15:17:40 Quit antil33t ()
15:17:56 Join HBK [0] (
15:21:08 Join Tuplis [0] (
15:21:50 Quit linuxstb (Ping timeout: 245 seconds)
15:22:19 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
15:22:35 Quit HBK (Read error: Connection reset by peer)
15:23:11 Join HBK [0] (
15:26:04 Quit HBK (Read error: Connection reset by peer)
15:26:09 Join KBH [0] (
15:26:54 Quit kugel (Ping timeout: 265 seconds)
15:27:11S_a_i_n_thow often does "viewers.config" change...realistically? Despit the "don't edit viewers.config" thing, I have, and wonder how often I need to update/or bother to check to update it.
15:27:46TheSevenwhy do you need to edit it in the first place?
15:28:13gevaertsit's included in the build zip, so "every time you install a new build"
15:28:26S_a_i_n_tto get viewers icons working in the "Open With list"
15:28:31S_a_i_n_t...a cheap workaround
15:29:50S_a_i_n_tviewer icons in "Open With" are applied incorrectly, or not at all...editing viewers.config is a cheap fix for that.
15:30:10TheSevenS_a_i_n_t: according to SVN, it is changed every ~2 months average
15:30:15 Join HBK- [0] (
15:30:32 Nick HBK- is now known as HBK ( "not very often" or, "when/if I notice something breaks"
15:31:03TheSeventhe changes are usually just new plugins
15:31:39S_a_i_n_tyeah, I'm probably the only one with icons for Frotz :D
15:32:36 Quit HBK (Read error: Connection reset by peer)
15:32:43 Join HBK [0] (
15:33:04 Join Rob2222 [0] (
15:33:25 Quit KBH (Ping timeout: 258 seconds)
15:33:41TheSeventhe last change of it was the addition of Frotz, btw :-)
15:34:43S_a_i_n_teven though I use "open with" very rarely, I *hate* seeing things in there with black squares or no icons at I bypassed that untill someone with brains finds a propper/the correct/ fix.
15:34:49 Quit HBK (Read error: Connection reset by peer)
15:35:04S_a_i_n_tthat's one of the things that made me notice it :D
15:35:14S_a_i_n_t*frotz that is
15:35:32 Join HBK [0] (
15:35:54CIA-88New commit by 03FlynDice (r24582): Sansa Clip+: Unset B0 correctly in dualboot.S
15:36:18S_a_i_n_tNot that I don;t have brains myself, I'm just not a coder...I could probably have a stroke and be a better coder than I am now.
15:36:38TheSeventhen do something about it :-)
15:36:50S_a_i_n_twhat? have a stroke?
15:36:57 Quit HBK (Read error: Connection reset by peer)
15:37:00*S_a_i_n_t knew what you actually meant :D
15:37:01TheSevenwell, if that helps? ;-)
15:37:38S_a_i_n_tOne of my problems is trying to learn too many languages (coding) at once
15:37:47S_a_i_n_tI should just stick with C
15:37:56S_a_i_n_tbut I get distracted VERY easily
15:37:57 Join HBK [0] (
15:38:05*TheSeven shoves that discusson to #rockbox-community
15:44:30 Join vegtoruci [0] (
15:46:56 Quit HBK (Read error: Connection reset by peer)
15:47:25 Join HBK [0] (
15:48:14 Join evilnick_B [0] (~0c140464@rockbox/staff/evilnick)
15:48:46 Join KBH [0] (
15:50:28 Quit KBH (Read error: Connection reset by peer)
15:51:02 Join KBH [0] (
15:52:12 Quit HBK (Ping timeout: 258 seconds)
15:52:33 Quit evilnick_B (Quit: Page closed)
15:55:01 Quit KBH (Read error: Connection reset by peer)
15:57:46domonokyis someone here who understands the subtrack handling in nsf ?
15:58:40domonokyi understand that it switches subtracks on trackskips if you are in repeat-one mode. but there is also internal playlist handling which i dont understand.
15:58:56domonokyi want to use the same/similar subtrack handling for the asap codec.
16:03:22***Saving seen data "./dancer.seen"
16:03:23 Join evilnick_B [0] (~0c140464@rockbox/staff/evilnick)
16:03:26 Join HBK [0] (
16:04:33 Join LambdaCalculus37 [0] (~rmenes@rockbox/staff/LambdaCalculus37)
16:06:06CIA-88New commit by 03teru (r24583): time menu: stop scrolling before leave the screen.
16:21:01 Join kugel [0] (~kugel@rockbox/developer/kugel)
16:22:15Unhelpfulsaratoga: i assume you mean *testing* on beast? also i'm still not really clear on how you would be able to use the 16x16 and 16x32 multiplies without losing accuracy... surely multiplying by constants truncated at 16 bits will introduce errors at around that precision
16:24:36Unhelpfulthere's a ? missing somehow... the other thing you might want to know about is that if you're making any use of smull or smlal and discarding the bottom word of the result, armv6 can do that without producing the low word in a register, and with the same throughput and latency as mul/mla
16:24:51 Quit grndslm (Remote host closed the connection)
16:25:25 Quit Zarggg_ (Read error: Connection reset by peer)
16:34:16 Quit AndyI ()
16:37:21 Join captainkewlllll [0] (~2669ecc2@gateway/web/freenode/x-sfhkmlohslcqebov)
16:41:53 Quit Tomis (Ping timeout: 252 seconds)
16:42:48 Join AndyI [0] (~pasha_int@
16:48:01 Join Tomis [0] (~Tomis@
16:48:23 Join panni_ [0] (
16:53:54 Quit Tomis (Read error: Operation timed out)
16:54:13 Join Tomis [0] (~Tomis@
16:55:11 Part watto
16:59:13 Quit Bagder (Quit: It is time to say moo)
17:01:03 Join watto [0] (~watto@
17:01:28 Join jgarvey [0] (
17:15:11 Quit einhirn (Quit: Miranda IM! Smaller, Faster, Easier.
17:19:21 Part evilnick_B
17:19:24 Join evilnick_B [0] (~0c140464@rockbox/staff/evilnick)
17:20:15 Quit Zagor (Quit: Clint excited)
17:25:44 Quit AndyI ()
17:26:00 Quit robin0800 (Remote host closed the connection)
17:31:36 Join AndyI [0] (~pasha_int@
17:32:15 Quit S_a_i_n_t (Quit: [St.] has exited mIRC™)
17:42:24 Part watto
17:45:03*TheSeven might have spotted why the channel swapping is happening
17:46:32TheSevenwill a pcm_more_callback_type always return an even number of samples?
17:46:45TheSeven(i.e. always start with the left channel)
17:47:26 Join watto [0] (~watto@
17:47:33 Quit linuxstb (Ping timeout: 265 seconds)
17:48:17 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
17:49:36 Join pamaury [0] (
17:53:16 Quit linuxstb (Ping timeout: 246 seconds)
17:53:45 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
17:54:55 Quit Sajber^ (Read error: Connection reset by peer)
17:55:21 Join Sajber^ [0] (
17:57:02 Quit teru (Quit: Quit)
18:01:28 Quit n17ikh (Ping timeout: 260 seconds)
18:03:24***Saving seen data "./dancer.seen"
18:04:44 Quit petur (Quit: work->home)
18:06:09 Join n17ikh [0] (
18:10:20 Join komputes [0] (~komputes@ubuntu/member/komputes)
18:19:40 Quit HBK ()
18:21:49 Quit AndyI ()
18:23:27 Quit togetic (Quit: WeeChat 0.3.0)
18:24:04 Join saratoga [0] (~9803c6dd@gateway/web/freenode/x-uxeftoofekqjpbkk)
18:24:14 Join togetic [0] (~togetic@unaffiliated/ibuffy)
18:25:28 Quit togetic (Client Quit)
18:25:42saratogalinuxstb: without dual core (AMS, Gigabeat, Nano2G) libmad is probably slower then libfaad
18:25:55saratogabut yes, I complain about libfaad a lot ;)
18:26:04 Join togetic [0] (~togetic@unaffiliated/ibuffy)
18:27:14saratogai suggested this as a possible GSOC project:
18:30:51 Join HBK [0] (
18:33:55 Join AndyI [0] (~pasha_int@
18:37:22 Join desowin [0] (~desowin@atheme/member/desowin)
18:37:39 Join bertrik [0] (
18:38:02topikandroid seems to have grown a bit since you created that gsoc ideas page
18:38:51 Join Zarggg [0] (~zarggg@2001:0:4137:9e74:0:4388:beb1:ba3d)
18:41:45 Join petur [0] (
18:41:45 Quit petur (Changing host)
18:41:45 Join petur [0] (~peter@rockbox/developer/petur)
18:42:47bertrikAlexP, nice to see your clip battery bench
18:43:03AlexPbertrik: Yeah, it wasn't bad
18:43:54AlexPSeems odd that you and I got much higher than the other one there
18:46:54bertrikyes, indeed, especially because runtime seems to be normal in the OF
18:47:13bertrikI'm trying to look into that, but it's a bit hard to get a grip on it
18:47:31 Quit swilde (Quit: ERC Version 5.3 (IRC client for Emacs))
18:49:44AlexPyeah, I can imagine
18:49:45 Quit archivator (Ping timeout: 245 seconds)
18:50:54 Nick YPSY is now known as Ypsy (
18:52:42topikbertrik: are you still looking at meizu's these days? i have an unused m6sl that would love to get a new life like my nano2g did after getting rockbox
18:53:36bertriktopik, I'm not really active lately for the meizus. I do have a M6SP and a borrowed M3, but not an M6SL that I can work on
18:54:32topiki read on the forum thread you had quite some progress
18:55:01bertrikthe main firmware for the m6sl probably does not compile, but you can probably get the bootloader compile with a limited amount of work, the display does not work at all yet AFAIK
18:55:30topiki tried to compile the bootloader and meizu_dfu but both didn't
18:59:27topikfair amount of m6sp stuff in the s5l8700 files, but not much m6sl
19:04:55 Quit bluebrother (Disconnected by services)
19:04:56 Join bluebroth3r [0] (~dom@rockbox/developer/bluebrother)
19:05:12 Join DumbSpammer [0] (
19:06:18 Join Casainho [0] (~chatzilla@
19:09:21Mode"#rockbox +o gevaerts" by ChanServ (ChanServ@services.)
19:09:32Kick(#rockbox DumbSpammer :you know why) by gevaerts!~fg@rockbox/developer/gevaerts
19:10:02Mode"#rockbox -o gevaerts" by ChanServ (ChanServ@services.)
19:10:22AlexPgevaerts: No ban?
19:10:36saratogaBagder: would you mind deleting that from the logs?
19:10:42TheSevenhe probably won't come back anyways
19:10:43AlexPA good idea
19:10:46gevaertsAlexP: I don't know. If he comes back, yes
19:10:57gevaertsand yes, I think that should be removed from the logs
19:11:02AlexPTheSeven: I know, but principle - if you come in just to spam then instant ban
19:11:19gevaertsAlexP: feel free
19:11:23AlexPnah :)
19:11:34TheSevenas he is probably doing that on a lot of channels, one might want to contact freenode staff
19:13:16saratogaB4gder: you too
19:21:12gevaertstopik: meizu_dfu builds fine here
19:21:29topikfrom the dir it is in? it complains about missing usb.h for me
19:21:43 Quit FlynDice (Remote host closed the connection)
19:21:56bertrikyou're missing libusb-dev then
19:22:35topikgood point. it compiles rightaway now
19:22:46 Join JdGordon_ [0] (~836b006a@gateway/web/freenode/x-iqqcwanrpzgvehlr)
19:27:21 Join Strife89 [0] (~michael@
19:27:27 Quit kugel (Quit: exit(0);)
19:28:00 Quit pamaury (Ping timeout: 264 seconds)
19:30:30 Quit togetic (Ping timeout: 265 seconds)
19:31:08 Join kugel [0] (~kugel@rockbox/developer/kugel)
19:41:52 Nick Ypsy is now known as YPSY (
19:42:49CIA-88New commit by 03b0hoon (r24584): Packard Bell Vibe 500: Improve/fix scrollstrip scrolling. The idea was taken from the ipod's clickweel source.
19:45:35 Join piotrekm [0] (~piotrek@unaffiliated/piotrekm)
19:46:06TheSevenhi piotrekm
19:46:26TheSevenif you have time, i would like you to test some things regarding the lcd issue...
19:46:35 Join Tomers [0] (
19:46:44piotrekmi might, i suppose
19:47:16piotrekmone quick question before: is the "tsst" (short white noise) on the beggining of track playback a nano2g-specific issue?
19:47:22 Join fml [0] (
19:47:43fmlJdGordon: ping
19:48:14 Quit Farthen (Disconnected by services)
19:48:22piotrekmTheSeven: any special instructions for today, or should I just re-download the files?
19:48:35 Join Farthen_ [0] (
19:48:37 Nick Farthen_ is now known as Farthen (
19:49:05TheSeveni would first have a check if this also happens when the lcd init code is run through ibugger
19:49:06 Quit vegtoruci (Ping timeout: 240 seconds)
19:49:11TheSeventhat would ease debugging
19:49:38piotrekmok, so first i'll reboot to linux
19:49:42TheSeveni'll prepare an image for you soon, need to fix something first
19:50:22fmlJdGordon: (for the log) if, as we know now, the font cache only stores the glyph codes and not the glyphs bitmaps, wouldn't it be better to just have one glyph cache for all fonts? Why do we need separate caches?
19:50:44JdGordon_well yeah
19:50:48JdGordon_although, maybe not
19:51:02JdGordon_I origionally thought it does store the glyphs bits which is why I did it this way
19:51:10piotrekmTheSeven: Anyway, I'll be back in about 40minutes.
19:51:19 Quit piotrekm (Quit: piotrekm)
19:51:32JdGordon_with more than one font loaded, its likely that their used glyphs wouold be different
19:51:57JdGordon_one mighyt only be used for numbers, anther might only be used for text
19:52:10JdGordon_and then there's the possibility of windings :p
19:52:16pixelma <- interesting picture as it shows a daughterboard with an "M5" marking. I suppose it's an X5V then (without the radio) but don't know if the loose wire has to so with it and if it should be this way
19:52:32fmlJdGordon: it's not unlikely. But on the other hand, it's very likely that the sets will be roughly the same. Chars are used for displaying text after all.
19:53:02 Join vegtoruci [0] (
19:53:17JdGordon_yes, so I might go back and simplify that completly
19:53:34JdGordon_probably change it to only cache the user font and not bother at all with the skin fonts
19:54:17fmlJdGordon: I think we can cache all fonts, no need to handle the UI font in a special way.
19:54:54CIA-88New commit by 03b0hoon (r24585): Packard Bell Vibe 500: Enable ATA DMA as in r24405.
19:55:00JdGordon_are you talking about the .glyphcache file? or the actual in ram cache?
19:55:09CIA-88New commit by 03kugel (r24586): Bubbles: Don't save scores when quit without saving is selected and reduce splash duration
19:57:31fmlJdGordon: both
19:57:33 Join FlynDice [0] (
19:57:36kugelwhat is gained by combining the glyph cache? combining means being potentially wrong and destroying the per-font glyphcache
19:57:51kugel(I mean the .glyphcache files)
19:57:52 Quit FlynDice (Remote host closed the connection)
19:58:30JdGordon_the on disk cache is tiny, so the only reason to not use a seperate one for each font is to cut down on junk files
19:58:46JdGordon_the inram cache NEEDS to be seperate because the bits are actually stored there
19:59:47fmlJdGordon: aha! Then we need separate caches indeed. Disk space doesn't matter
20:00:01kugelthe inram cache could be combined
20:00:21JdGordon_it can, but the gain isnt worth it
20:00:27 Join archivator [0] (
20:00:27kugelit just needs a font index in the lru nodes, and it actually has an unused field
20:00:39kugelisn't worth what?
20:00:47JdGordon_that doesnt work when each font has different glyph sizes
20:01:04JdGordon_the wait and the effort and the extra ram usage
20:01:06fmlkugel: the entries would be of variable size which is not supported now
20:01:36fmlkugel: since the bits are also inside
20:01:58 Quit evilnick_B (Quit: Page closed)
20:02:00kugelit is supported
20:02:23fmlkugel: I don't think so
20:02:23kugelthe lru node has a data pointer and a size field
20:02:57gevaertsarchivator: I'm test-compiling things now. There are a few keypad issues that i need to look into
20:03:22JdGordon_kugel: ok, if you really want it done that way then go for it, in the mean time, my way works, and is almost ready, and we dont have to deal with issues of dropping X glyphs to make room for Y other fonts glyphs
20:03:27***Saving seen data "./dancer.seen"
20:03:45fmlkugel: and how is that data managed? It has to be of a const size
20:04:10kugelJdGordon: yea, but with constant disk spinning because 10k seems pretty low
20:04:34JdGordon_(I havnt tested on target yet)
20:04:52JdGordon_10k is how many 12x12 glyphs?
20:05:08JdGordon_1000/(12x12/8) ?
20:05:12gevaerts70 or so
20:05:34JdGordon_thanks, so easily enough for most people's secondary font
20:05:50gevaerts55 actually
20:06:27kugelfml: why?
20:06:41rasherThat's not a lot
20:07:08JdGordon_rasher: for the secondary font? its plenty
20:07:17rasherHow do you figure?
20:07:18JdGordon_even if its closer to 45, still plenty
20:07:21fmlkugel: look at the code. It's handled as an array, i.e. fix sized entries
20:07:25amiconnAhem. 10k would be 10000
20:07:48rasherIt's entirely plausible that the secondary font will be needing the same glyphs as the main font
20:08:01amiconnBut then, plain /8 is incorrect either, as the bitmaps cannot contain fractional bytes
20:08:04gevaertsok, 568
20:08:05kugelI can imagine the user fonts are used for bigger fonts so that the menu font isn't too huge, then 12px is a bad estimate
20:08:07TheSevensomeone broke nano2g playback!
20:08:25gevaertsamiconn: that would mean (16x12)/8?
20:08:39gevaertsok, 426 then
20:09:00JdGordon_10k is a number I pulled fairly randomly, we can increase that if its feeled we need to
20:09:10JdGordon_ok 400 glyphs IS plenty
20:09:23gevaerts140 glyphs for a 24x24 font
20:10:06kugelI wouldn't underestimate the overhead from the cache management either
20:10:07fmlThat's too much IMHO
20:10:38kugelit could probably be 10% less due to that
20:10:39JdGordon_kugel: with a few lines of code we can get exact numbers
20:10:52JdGordon_10% of 426 is still only 42
20:11:03JdGordon_worst case is 300
20:11:06JdGordon_or 100
20:11:29JdGordon_fml: the RAM being used for fonts is preallocated for skins anyway so its not a waste
20:12:02fmlJdGordon: ok then
20:13:16JdGordon_has anyone looked at the latest patch? should I simplify it and move the skin_font_*() stuff into firmware/fonts?
20:14:04JdGordon_the only downside doing that is a fwe extra font structs in firmware/ but it should make the code ismpler
20:14:35 Quit amiconn (Disconnected by services)
20:14:37 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
20:14:45 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
20:14:45 Quit pixelma (Disconnected by services)
20:14:59 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
20:15:00kugelfml: you seem correct
20:15:02 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
20:15:24TheSevenCC apps/codecs/asap.c
20:15:26TheSevenmake: *** No rule to make target `/home/theseven/rockbox-trunk/apps/codecs/libasap/anylang.h', needed by `/home/theseven/rockbox-trunk/build/nano2-app/apps/codecs/libasap/acpu.o'. Stop.
20:15:35JdGordon_make veryclean
20:17:15 Quit archivator (Ping timeout: 245 seconds)
20:18:24fmlJdGordon: I wouldn't handle fonts diferently at the firmware level. Make them equal. A font just gets a buffer for storing its cache. The caller must care about the buffer. Then the font buffer can be in the skin buffer or in the playback buffer (if we want it that way) or...
20:19:22amiconnImo the proper solution would be a single font buffer for all fonts
20:20:01fmlJdGordon: one font is special in the sense that it's the UI font and has its buffer in the core rockbox. Other fonts get the buffer from the skin. This can be changed later without changing the firmware.
20:20:04kugelI agree, but nobody is going to do it, and JdGordon's approach seems good enough
20:20:33*amiconn disagrees
20:20:58amiconnAt the firmware level, all loadable fonts should be equal
20:21:00fmlkugel: I think with the energy and speed JdGordon has it will be done in two, no, one hour
20:21:12JdGordon_fml: so yeah, thats how I've done it now, but its got a vit too much special handling for the ui font
20:21:40amiconnThis is important on devices with more than one display
20:22:02fmlamiconn: you mean all but the system (inbuilt) font? Then I fully agree.
20:22:03JdGordon_I agree that one buffer is better, but noone has done it and looks like they are going to
20:22:12*JdGordon_ totally forgot about multiple screen fonts :p
20:22:16amiconnEach display should be able to set its own ui font. That's in fact what's annoying me most
20:23:07amiconnI don't really need more than one font per display, but the single font for all lcd gets annoying, especially if there's a large difference in lcd size
20:23:37amiconnThink e.g. H300 main and remote, or even more obvious m:robe 500 and remote
20:23:39fmlamiconn: ..which is always there
20:25:14amiconnJdGordon_: markun said that he did a bit of work on the font caching. Unfortunately he didn't get around to committing it
20:25:15JdGordon_so I'm tempted to duplicate the user font handling for the remote (and just make sure the right FONT_UI is loaded for the display if the fonts are different), that wll double the buffer usage
20:26:54Unhelpfulone buffer pretty much means coming up with a way to get entries of differing size into it
20:27:24JdGordon_which is what has been the problem the whole time
20:27:35JdGordon_also coming up with a good buffer size to use
20:27:45amiconnThe single buffer method has advantages if the theme has some special-use fonts
20:27:49 Join moos [0] (moos@rockbox/staff/moos)
20:28:15Unhelpfuli'd say maybe something like what buflib does - there's a table of offsets to the actual buffer items (quick access, one extra indirection vs direct array lookup) but the buffer items themselves are of arbitrary sizes.
20:28:27amiconnThink e.g. of an fm theme for users not caring about the station name. It would use a huge font for frequency display
20:28:37Unhelpfulactually it's not an offset table, but a pointer table
20:28:55amiconnOf that huge font, only the digits, '.' and maybe 'M', 'H' and 'z' are needed
20:29:06JdGordon_Unhelpful: that doesnt solve the problem of figuring out which glyphs to drop to make room for a different one
20:29:48amiconnUnhelpful: The pointers also cost 4 bytes extra per entry. But sure, it would probably already be an overall gain with a medium-sized proportional font
20:29:53JdGordon_i.e you dont want to just drop the oldest one which is larger than the needed space, that one might be really really frequently used
20:30:17Unhelpfulamiconn: that's what i'm thinking, a proportional font will rather quickly save when it caches narrow characters
20:30:21amiconnJdGordon_: You just drop from oldest to newest until you have enough free space
20:30:32 Join stooo [0] (
20:31:13JdGordon_and the buffer size? 60K is obviously too small
20:31:31Unhelpfulbuflib could also be reworked a bit to support unaligned data - right now it's 32-bit aligned so 1) users never have to worry about allocation alignment 2) each item has a pointer to its table entry for easy update on compaction
20:31:32JdGordon_100K? probably too small for people that want 3 fonts, too big for those that only want one
20:32:54JdGordon_60K is how much we allocate now for one font, so it stands to reason that it's not enough for 2+
20:32:59fmlamiconn: and after each drop you compact the buffer?
20:34:12amiconnYes, I think that's necessary
20:34:36JdGordon_wont that get very slow? doing lots of memcpy?
20:34:40gevaertsJdGordon_: I'd assume that if people use more than one font, in many cases the extra fonts will be used for only a few words
20:34:41amiconnYou might also track gaps and only compact if no large enough gap is left, but that tracking might be too much overhead
20:35:11amiconnJdGordon_: It needs memmove. Compared to the disk access, this memmove is blazingly fast
20:35:17JdGordon_gevaerts: agreed, which is why I did 10K not the full 60K
20:35:29amiconnBtw, on lowmem swcodec the font buffer is 10k, and on archos even just 4k
20:36:01amiconnI would make that depend on lcd size. Larger lcds are likely to need larger fonts, hence a larger font buffer
20:36:07fmlSo we'd need yet another mini malloc
20:37:36Unhelpfulamiconn: that's what buflib does already, compact-on-failure. also handles are allocated from the buffer so that there's not a fixed number of handles.
20:37:59 Join phanboy4 [0] (
20:38:13 Join archivator [0] (
20:38:41Unhelpfulsearch for allocation is done by walking the buffer items - since they each have an allocation size in their "header" they can be treated as a linked list. as an optimization a pointer to the lowest free buffer is maintained.
20:38:57JdGordon_so is this just all talk? if someone is actually going to do this then I'm happy to sit on my patch, but if its just going to stay talk (like it has been for years) then I dont see any reason to wait once my patch is ready
20:39:24gevaertsarchivator: I think I'll disable fft on touchscreen targets for now
20:39:50amiconnUnhelpful: Hmm. How does that perform if there are 1000 items or so?
20:40:20amiconnAnd does it reuse gaps?
20:40:22 Join piotrekm [0] (~pm@unaffiliated/piotrekm)
20:40:49amiconnJdGordon_: Imo getting that right is a requirement for multifont
20:41:39Unhelpfulamiconn: it depends on how out-of-date the gap information is, and i've never really tried to benchmark or stress-test it. but the lower-bound pointer for a free buffer is updated on each allocation. the worst case is if an allocation uses all of the lowest-address gap, since this means that the lower-bound pointer will actually point to a non-gap
20:41:43archivatorgevaerts: I was actually considering dropping the current oscilloscope-inspired keymaps and try menu-based configuration... Shouldn't be too much work but I'm afraid it might make the plugin less usable
20:41:46JdGordon_I agree that a single buffer is better, but this same conversation happens at least once a year and there has been no movement
20:41:52 Join fleebailey33 [0] (fleebailey@unaffiliated/fleebailey33)
20:41:54piotrekmhello again
20:42:05piotrekmTheSeven: no i'm availabe
20:42:11gevaertsarchivator: I'm about ready to commit, with a note that keymaps need more work
20:42:13Unhelpfuli primarily optimized it for fast retrieval rather than allocation, but the allocation should be quite speedy in the typical case
20:42:26gevaertsTell me if I should wait
20:42:32*amiconn thinks we need more devcon-like hacking sessions
20:43:07archivatorgevaerts: no, by all means, commit it. I don't have time to do any serious refactoring right now... menu-based config was just a thought..
20:43:32archivatorgevaerts: oh, and thank you :)
20:44:05Unhelpfulthe same optimization is used in the handle table, so that allocating a free handle starts with the lowest possible handle number
20:44:20CIA-88New commit by 03gevaerts (r24587): New plugin: FFT, A frequency analyzer plugin ...
20:45:52gevaertsarchivator: if you could work on a manual patch, or at least a bit of text, that would be very welcome!
20:46:23kugeleuw, typedefs :(
20:46:33archivatorgevaerts: I'll see what I can do. I've never touched LaTeX though :(
20:46:59gevaertskugel: imported code
20:47:35gevaertsarchivator: we probably need a line in CREDITS for that external code
20:47:46kugelif it's imported, why are there no (c) headers?
20:48:59gevaerts"Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution"
20:49:26archivatorwhere are you seeing that?
20:49:41gevaertsok, standard BSD apparently
20:49:52archivatoryes, it is.
20:51:05kugelarchivator: did you lack a malloc()? we have one for codecs and plugins
20:52:03archivatorkugel: I don't remember ever needing a malloc but I could be forgetting something. What do you have in mind?
20:52:29gevaertsHow do we handle credits for single-person projects where the author didn't contribute directly? The main name list, or the team list at the end?
20:52:34kugelI saw some #define *_MALLOC mallic and #define *_FREE free, so I just though
20:53:16 Quit flydutch (Quit: /* empty */)
20:53:31kugelgevaerts: I think we add them to the people's list, that's what happened with the timestretch feature
20:54:16kugelif someone doesn't want to be named he shouldn't release code to the public :)
20:54:21CIA-88New commit by 03gevaerts (r24588): Add Mark Borgerding to CREDITS, for kiss_fft which is used by the fft plugin
20:54:54gevaertswell actually, the license says we have to add him if I read it correctly :)
20:55:14archivatorkugel: I didn't put too much effort into cleaning that library. I think the OPENMP stuff is still there..
20:56:05 Part watto
20:56:15kugelI also saw "this kiss fft can't use malloc" in a .c file
20:56:38archivatorkugel: Ah, yes. I remember now. It's not that it can
20:56:47archivator't but rather that I didn't want it to.
20:57:48archivatorthe library in its current form does not depend on core rockbox but only on the fp_sincos implementation. i wanted to confine all dependencies on core rockbox to the actual plugin.
20:58:13kugelour malloc isn't core :p
20:59:38amiconngevaerts: Could you quick-check something on your D2?
20:59:57amiconn(APE, testing just -c2000 and -c3000 is sufficient)
21:00:01archivatorkugel: then I must have been seriously confused :) In any case, malloc isn't really needed - the allocation was made static as the size of the transform was fixed at compile time
21:00:25amiconnI want to make sure the armv5te does actually behave as described before doing more complex stuff
21:00:33amiconnWant a patch or a build?
21:00:36kugelwell, your patch was maybe longer on the tracker than the time we have it
21:00:39gevaertsa patch is fine
21:01:12piotrekmgevaerts: is this fft plugin (it's great tirst of all) meant to have some config, refresh time for example? i wonder if it's only because of the unbound keys
21:01:13 Quit TheSeven (Read error: Connection reset by peer)
21:01:30kugelpiotrekm: ask archivator
21:01:35archivatorpiotrekm: no, not really. It refreshes as often as it gets data.
21:01:37gevaertspiotrekm: archivator can answer all questions about it!
21:01:40 Join TheSeven [0] (~theseven@rockbox/developer/TheSeven)
21:01:54 Quit Casainho (Quit: ChatZilla 0.9.86 [Firefox 3.6/20100115133306])
21:02:06kugelI wonder if it can also analyize line in stuff?
21:02:09 Quit fml (Quit: CGI:IRC)">
21:02:55amiconngevaerts: With this patch, APE should decode at the same speed or a little bit faster than SVN on D2
21:03:11amiconn(a bit faster due to more registers left for gcc to play with)
21:03:21piotrekmTheSeven: in case the last message didn't arrive, i have some time now to test
21:03:24 Quit linuxstb (Ping timeout: 264 seconds)
21:03:48archivatorkugel: in theory, yes. I'm not sure how to switch between pcm output and line in mode. Can you record and have something playing at the same time? Sorry for the silly question, I've never recorded anything on a DAP..
21:06:57TheSevenpiotrekm: i uploaded a file called ibugger.bin to my server, can you check if the lcd driver used by that works?
21:07:40 Quit stooo (Ping timeout: 245 seconds)
21:07:47TheSevenif yes, get inithw.bin and run sudo python upload 22000800 inithw.bin && sudo python execute 22000800 22000800
21:08:04TheSevenrun that within ibugger *loader*, so don't run startup before
21:11:32gevaertsamiconn: playback is still fine, test_codec results as soon as I stop doing things wrong
21:11:53TheSevendamn, just had another FTL crash
21:12:15amiconnYou don't need to test correctness, I'm doing that on my beast (using the armv5 vector math of course)
21:12:37piotrekmTheSeven: see a mandelbrot;)
21:12:49TheSevensounds like the driver works
21:12:52 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
21:12:56TheSevenwhat does happen if you run inithw.bin?
21:13:07piotrekmwill do that in a moment
21:14:10piotrekminithw gave white screen
21:18:17gevaertsamiconn: c2000 317%, c3000 213%
21:18:22 Join anewuser [0] (~anewuser@unaffiliated/anewuser)
21:18:47amiconnAh, very good. So armv5 actually behaves as described :)
21:19:03*amiconn starts working on the actual improvements
21:19:31 Quit kugel (Remote host closed the connection)
21:21:50 Quit linuxstb (Ping timeout: 245 seconds)
21:23:48saratogaarchivator: we're about to add a very well optimized FFT to the library FWIW
21:24:46archivatorsaratoga: I'm aware of that, I'm waiting for it to go upstream and I'lll switch to it (if it's meant to be used by plugins, that is).
21:25:43 Quit Tomers (Quit: ChatZilla 0.9.86 [Firefox 3.5.7/20091221164558])
21:25:44saratogaok cool
21:25:51saratogaits in the mdct branch if you'd like to try it
21:26:27TheSevenpiotrekm: hm, not good
21:26:39TheSevenbut at least we can test it easily now
21:26:43archivatorsaratoga: is the api stable? I.e., can I start switching to it now?
21:26:51piotrekmTheSeven: does it mean something else than lcd init is broken?
21:27:29TheSevenno, it probably is the lcd init, and we can now try to track the issue down
21:28:47 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
21:40:14saratogaarchivator: its not even exposed to plugins at the moment, but if you add it to plugins.h I doubt anyone will change it now
21:40:41saratogaits actually just a function used by our new MDCT, so we didn't need to expose it
21:40:55saratogai wouldn't worry too much, we can always do this later
21:41:32archivatorRight. I might work on this when I find the time (uni has been proving somewhat time-consuming :) )
21:41:44 Quit FOAD (Ping timeout: 260 seconds)
21:42:08saratogashould be quite simple unless your fft is doing something special (reordering or scaling)
21:43:12 Join FOAD [0] (
21:45:07archivatorsaratoga: it's doing scaling at the end of each iteration to prevent overflow. Other than that, it's pretty much taken from the books verbatim.
21:47:15TheSevenmeh. i don't even get my nano to play back *anything* any more
21:47:24 Quit piotrekm (Ping timeout: 246 seconds)
21:49:01 Join piotrekm [0] (~pm@unaffiliated/piotrekm)
21:49:46piotrekmTheSeven: you mean it's permanently broken?
21:50:18TheSevenno, looks like somebody broke a bunch of things with a recent commit...
21:50:29TheSevenand it's reacting really weird
21:50:45TheSevenright now it failed to rolo a new build because of a checksum error?
21:51:05TheSevenoops, and there suddenly is a file called "tagna6i.conf)g"
21:51:09piotrekmi have the latest commit on mine and it works fine, with mp3s at least
21:51:12TheSevenerm, wtf...
21:51:41TheSevenwell, i have enabled undervolting and boosting, but that was working fine before...
21:51:46piotrekm*not latest, 24587
21:51:53UnhelpfulTheSeven: disk corruption much?
21:53:09 Join mt [0] (~mtee@rockbox/developer/mt)
21:55:22 Quit Strife89 (Quit: Going home.)
21:56:55TheSevenUnhelpful: it also keeps prefetch/data aborting
21:57:55 Join froggyman [0] (
21:59:32 Quit phanboy4 (Quit: Leaving)
22:00:17 Quit ender` (Quit: 'And you have to shout -' He tried to remember some far-off reading. '- er, bonsai. Yes. Bonsai!' -- Terry Pratchett: Reaper Man)
22:03:31***Saving seen data "./dancer.seen"
22:14:23TheSevenargh, that channel swapping issue is driving me mad
22:14:41TheSeveni have files where it seems to swap like 5 times a second, and others that don't swap at all
22:20:55CIA-88New commit by 03theseven (r24589): Fix iPod Nano 2G channel swapping issues
22:25:28 Quit archivator (Quit: Leaving)
22:25:32liarthanks TheSeven
22:28:02piotrekmthanks TheSeven
22:29:56 Join Strife89 [0] (
22:33:40TheSevenwell, that was a trivial one
22:37:09piotrekmTheSeven: you know that it also fixed the white noise appearance on the beginning of tracks?
22:37:41piotrekmor not... i said that too early
22:37:43 Join Jaykay [0] (
22:38:42 Quit JdGordon_ (Ping timeout: 248 seconds)
22:38:44piotrekmit's better than was anyway;)
22:39:04AlexPTheSeven: Can FS #10812 be closed?
22:40:33 Quit captainkewlllll (Quit: Page closed)
22:40:49 Join Buschel [0] (
22:41:21TheSevenpiotrekm: i never have noise at the beginning of tracks...
22:41:43TheSevenAlexP: thanks, was digging through flyspray right now to find it
22:41:58 Quit mt (Remote host closed the connection)
22:42:06TheSevenwe really need a way to list all bugs for a given targetr
22:42:12 Join JdGordon_ [0] (~836b006a@gateway/web/freenode/x-xhpxrbgutegziryi)
22:42:19piotrekmi get it in some albums, but when played on pc this doesn't occur
22:42:39AlexPYeah - I was looking for that just now but it doesn't seem to appear in the advanced search bit
22:43:27TheSevenpiotrekm: i sometimes get parts of other tracks etc. played for an instant when skipping around
22:44:02AlexPB4gder: When you get a second could you put me so I can assign flyspray things to myself? I'm not in the list of people to assign things to and a couple of times I wanted to give myself things
22:44:30piotrekmTheSeven: i think i wasn't precise again, this occurs only when one track ends and another begins, not if i manually skip
22:45:06 Quit Omlet (Read error: Connection reset by peer)
22:45:13B4gderAlexP: checking...
22:45:47B4gderAlexP: so you have no "assign to me" button then?
22:46:00AlexPoh, er, maybe - one mo :)
22:46:43 Join Omlet [0] (
22:47:30AlexPB4gder: Sorry, yes - I didn't notice that and was trying to do it via the list when you edit the task :)
22:47:39 Quit Farthen (Ping timeout: 256 seconds)
22:47:48B4gderoh well, issue sorted then I guess! ;-)
22:47:52AlexPyep :)
22:48:12AlexPNot sure how I didn't spot that button - it is right next to close task which I use quite a bit :)
22:48:35B4gderyou've been SO focused on that close!
22:50:23TheSevenB4gder: is there any way to add the player type to the filter options in flyspray's advanced search?
22:51:53B4gdernot that I know of, no
22:52:26TheSeven(besides patching flyspray)
22:52:38 Join Adnyxo [0] (
22:55:55TheSevenheh, I *think* I may have found a way to get rid of those boosting glitches
22:56:19TheSevenwhat about just switching the ARM core to fastbus mode instead of tampering with the prescalers?
22:56:23 Join Horscht [0] (~Horscht2@xbmc/user/horscht)
22:57:13bertrikTheSeven, I think that's what we do for the as3525 targets
22:57:27 Join Strife89|PalmTX [0] (
22:58:08TheSevenwe may want to additionally clock down the system bus, but as soon as the cpu and the bus are using the same clock, I don't think the freqswitch will cause any issues
23:01:42 Quit Strife89|PalmTX (Client Quit)
23:03:57saratogaB4gder: what did the ffmpeg people want from us?
23:04:24B4gderwma, mdct and ape came up in the discussions
23:04:43B4gderI couldn't really comment on the details, as I don't know them...
23:04:46saratogado you have the logs?
23:04:55saratogai don't think they log their channel
23:05:14B4gderehm, I don't right now as I did it from my other computer which is shut off atm
23:05:19bertrikWe have a person called "Phinitnun Chanasabaeng" and another called "Pinitnun Shanasabang" in CREDITS. Are they actually different people?
23:05:44saratogai've given merging our code a lot of thought and basically decided it wasn't practical, so i'm curious what they were thinking exactly
23:05:50B4gdersaratoga: ask them, they're friendly and many in that channel ;-)
23:05:58saratogahaha thats not been my experience
23:06:32saratogalast time i asked the answer was something like "go away and stop forking ffmpeg"
23:07:05saratogaalthough in fairness they were talking about AAC, and we don't even use their aac decoder, so the guy probably just didn't know what he was talking about
23:07:13bertrikmaybe they didn't realise how serious you are about it
23:07:54saratogathe ffmp3 is a no brainer, and in fact its been proposed as a potiential GSOC project for this year
23:08:04saratogathe same for ffaac
23:08:16saratogaalthough they overestimate their improvements over libmad
23:08:25saratogai think i saw 15% on x86 over libmad
23:08:28B4gderI somehow suspsected that
23:08:39saratogawhich puts them somewhere around "astonishingly slow"
23:08:40B4gderthey claimed ~50%
23:08:55saratogayou sure they didn't mean libfaad?
23:09:09saratogathey're about 50% faster then libfaad and people mix up it with libmad sometimes since they rhyme
23:09:28B4gderwe explicitly talked about mp3 then so I'm quite sure
23:09:37saratogaFFmpeg: 22.3s libmad: 24.2s mpg123: 9.5s
23:10:00 Quit anewuser (Quit: Another edition of chiptune gig WinterChip5! :O =ooo)
23:11:31saratogaalthough they've obviously got some algorithmic improvements, so its worth seeing if we can steal them
23:11:56saratogado they actually want their fixed point mdct back?
23:12:18saratogai remember reasonably efficient fixed point mdcts being propose to them before and going no where
23:12:58saratogai assumed they weren't interested
23:13:08B4gderthey do, but I'm not sure for what purpose exactly
23:13:50TheSevenhm, fastbus saves ~4mA
23:13:55saratogamt's cook decoder is basically a dusted off set of patches for fixed point code in ffmpeg
23:14:01TheSevenand no glitches any more :-)
23:14:05saratogaTheSeven: the arm9 bus mode?
23:14:26bertrikwow, nice! current consumption was already pretty low on the nano 2g, right?
23:14:30saratogacombined with his own rm parser and a lot of optimization of course
23:14:40TheSeven21/22 vs. 25/26 mA
23:14:50saratogafastbus is where you run the clocks at integer multiples right?
23:15:21bertrikno, that's synchronous
23:15:31TheSevenfastbus = arm runs at bus clock
23:15:58TheSevensynchronous doesn't work for me, even when they are running at integer multiples
23:16:06saratogaso how do you boost?
23:16:17TheSevenby switching between asynchronous and fastbus
23:16:27saratogathats probably something we should check on ams then
23:16:33TheSeven(and additionally clocking the bus down to 50MHz when in fastbus mode)
23:16:45saratogahow does IRAM work on the 2G?
23:16:57saratogais it off the AHB or some special bus
23:17:03TheSevenprobably ahb
23:17:10TheSevenit needs 2 cycles per access IIRC
23:17:19saratogaisn't the AHB really slow
23:17:33saratogai thought its a 16 bit bus with multiple cycle access times
23:17:33TheSevenAHB is 100MHz, ARM is 200MHz
23:17:45saratogaso that'd be no better then 4 clocks
23:17:52TheSevenhm, iirc AHB is 32bit...
23:18:05 Quit _zic (Ping timeout: 245 seconds)
23:18:06TheSevenbut it may well be a different bus
23:18:16saratogaah ok
23:18:39saratogathe one on AMS is either amazingly slow, or else the IRAM is really slow
23:20:55bertrikthe as3525 calls AHB the "the advanced *high-speed* bus"
23:21:22saratogathe ARM9TDMI also says it has a fast 4 cycle multiplier
23:21:40saratogatrust nothing arm tells you
23:22:09bertrikI have no reason to believe it's slow
23:24:14TheSevenoops, my nano is starving for cpu power, the menu is *very* sluggish, but it just doesn't want to boost (stays at 47MHz)
23:25:01TheSevenbut it's still playing back realtime, only had 2 little gaps during that track (mp3)
23:26:59 Quit Jaykay (Quit: ChatZilla 0.9.86 [Firefox 3.6/20100115144158])
23:28:00 Quit GeekShado_ (Read error: Connection reset by peer)
23:29:39amiconngevaerts: Could you test a new patch? This time with real speed improvements...
23:30:57 Quit Omlet (Quit: ( :: NoNameScript 4.22 :: ))
23:32:19amiconnPlease test -c2000 and -c3000 speeds first. I expect around 7% speedup for -c3000. If that works, you could do a full test; this will be the final version except some #if 0'd junk
23:33:00 Quit piotrekm (Quit: Leaving.)
23:33:51*amiconn would be interested in the raw test_codec output as well, i.e. not rounded
23:35:55Unhelpfulsaratoga: well, average 4 cycles?
23:36:08*JdGordon_ points everyone to the mailing list
23:36:24JdGordon_pixelma: amiconn: any chance you can test the fm patch on hwcodec+rec soonish please?
23:36:46JdGordon_I'm at the pointt where its ready for commit
23:37:38gevaertsamiconn: c2000 220.01%
23:37:47gevaerts330.01 that is
23:39:25 Join matsl [0] (
23:39:40gevaertsamiconn: c3000 220.92
23:39:48JdGordon_rasher: domonoky: does anything need to be done for the theme site to accept fm skined themes? or will they just work (but not verify them) untill magic scripts are fixed?
23:40:16domonokyJdGordon_: it will accept them.
23:40:47domonokyto get them verify i just need to add extensions (if checkwps takes them without problems)
23:41:36JdGordon_checkwps shouldnt have any problems, I dont tihnk it checks extensions
23:41:43JdGordon_does the theme site verify the .cfg?
23:42:29 Quit Buschel ()
23:43:54gevaertsJdGordon_: just to clarify once more, "coming from the skin buffer" is talking about the skin usage, right? i.e. if it gets added to the plugin API, plugins would provide the RAM?
23:44:37JdGordon_gevaerts: right now, no plugins would use the skin buffer also
23:44:57JdGordon_but thats something that can change
23:45:13amiconngevaerts: That's in fact reasonable. When estimating, I didn't take the other stages into account
23:45:25amiconnSo filter speedup is ~9%, as it should be
23:46:02JdGordon_the problem with passing extra buffers in is what happens when a plugin loads font A, and something else also loads font A? right now A is only loaded once and not unloaded untill both unload it. but if the first buffer points into the plugin buffer it will break
23:46:10JdGordon_that shouldnt ever happen though
23:46:13amiconnCould you please do the other compression levels as well (including -c1000 - sometimes shifting code around changes speed due to cache influences)
23:46:56gevaertsamiconn: running now. It just started on c5000. I'll pastebin the output file once it's done
23:47:03gevaertsJdGordon_: right now it's not exported, so there is of course no issue :)
23:47:22domonokyJdGordon_: no it doesnt check the config at moment.
23:47:35 Join pamaury [0] (
23:47:48gevaertsI see your point, but what if a WPS uses the full buffer, and then a plugin starts? Does the WPS get reloaded afterwards?
23:47:52domonokythat is a todo for the future. is there are list of allowed settings for themes ?
23:48:04JdGordon_gevaerts: yes, but even when it is, the skins are the only things loading fonts, and that cant happen when a plugin is running (simply because it uses the plugin buffer to read the skin file :) )
23:48:13 Quit pamaury (Client Quit)
23:48:39 Join pamaury [0] (
23:48:48JdGordon_domonoky: no, but the list is fairly static so if its just a file in svn we can update it as needed
23:49:06gevaertsJdGordon_: ok, so a plugin can never fail to load a font because a the user happens to like complex wpses?
23:49:35JdGordon_it can now,
23:49:59JdGordon_yeah, so plugins do need to pass in a buffer somehow
23:50:09 Quit domonoky (Read error: Connection reset by peer)
23:50:49gevaertsI think so, yes
23:51:15gevaertsApart from that, I agree that a multi-buffer approach now is better than a nice single-buffer appoach in three years
23:52:17JdGordon_also, with a sufficiently complicated theme a plugin could fail to load a font because the theme uses up all the font slots
23:52:39JdGordon_unless we can come up with a better way of storing the font pointers
23:53:15gevaertsIf possible, I'd look for a way for plugins to have their own font slots and only share the two main ones
23:54:04gevaertsor we could restrict font slots per user
23:54:09JdGordon_so we use a callback for the lcd drivers to get the font from the plugin?
23:55:37gevaertshm, yes. The lcd drivers shouldn't really have to know about plugins...
23:56:27JdGordon_maybe we just up the font slots to something like 20 and assume noone will be able to fill that
23:56:38JdGordon_17 avialable for the apps/ layer ought to be plenty
23:57:23gevaertsI'd reserve a few explicitely for plugins then
23:57:43JdGordon_how many? 3? 4?
23:58:01*amiconn thinks this sounds more complex than a single font buffer
23:58:17 Quit pamaury (Quit: abort();)
23:58:39JdGordon_amiconn: yes, but unless someone actually says they are going to code it then this is what we have
23:58:45gevaerts3 or 4 is plenty, yes. The only plugin that we know really wants more fonts just needs one more after all
23:58:50JdGordon_also, seperate buffers means less likelyhood of cache thrashing

Previous day | Next day