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 2007-10-22

00:00:08amiconnNow if I knew what penalty branches have on arm7tdmi...
00:00:14 Nick bryan_ is now known as Toxicity999 (n=bryan@cpe-76-179-68-20.maine.res.rr.com)
00:00:46 Join pienose [0] (i=5614aab7@gateway/web/cgi-irc/labb.contactor.se/x-a9d0956a7cba800d)
00:01:17 Quit Mouser_X (Read error: 104 (Connection reset by peer))
00:01:29 Quit pienose (Client Quit)
00:01:30preglowBuschel: did you see i found out what was causing moos mpc corruption?
00:01:51Buschelpreglow: he told me yesterday about pre-amp. this correct?
00:01:54preglowyeap
00:02:01Buschel:o)
00:02:06 Join Mouser_X [0] (n=someone@207.155.176.3)
00:03:44Buschelamiconn: damn! it's much faster when always clamping!
00:03:52amiconnIs it?
00:04:10Buschelyes, goes up from 8.7 -> 10.4 fps
00:04:20*amiconn remembers having seen an arm instruction cycle time table somewhere, but can't remember where :/
00:04:46amiconnpreglow?
00:05:31 Quit grndslm (Remote closed the connection)
00:08:01 Quit kratonator ()
00:08:09Buschelhmm, my fault :/ it's faster on the test-image, but not that much. on real video it's a bit slower
00:08:41 Join grndslm [0] (n=grndslm@24-116-87-97.cpe.cableone.net)
00:08:46 Join japc [0] (n=japc@bl7-241-88.dsl.telepac.pt)
00:09:04preglowamiconn: i pasted you an url once, can't remember it now
00:09:21amiconnAha, probably because the test image clips in the outer parts
00:10:28 Quit desowin ("use linux")
00:12:27amiconnhttp://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.ddi0210c/Chdbedbg.html
00:12:27preglowamiconn: http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.ddi0210c/Chdbedbg.html
00:12:30*amiconn hugs grep
00:12:30preglow:)
00:12:36amiconnlol
00:15:07amiconnAh, branch is 3 cycles when taken, so that general clamp check needs 6 cycles in the standard case
00:18:17makuseruhow can i get all my songs organized by artist on my gigabeat, when i put it on there its all just in one folder and thats alot of scrolling to get to the botom, is there no way to organize by artist, than click on the artist and get to the abums and click on that to go to songs, etc etc, is there any way to do that
00:18:36 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
00:18:47bluebrotheruse some pc software to reorganize the files on the player or simply use the database
00:18:56makuseruhow can i do that?
00:18:58Buschelamiconn: but we save a register for holding the port address
00:19:05Mouser_Xmakuseru: Read the manual?
00:19:07*bluebrother points to the manual
00:19:19makuseruumm, what manual?
00:19:33Mouser_XThe manual linked on Rockbox's site?
00:19:40 Join safetydan [0] (i=dc9d468b@rockbox/developer/safetydan)
00:19:46amiconnBuschel: where?
00:19:48bluebrotherthe manual that is linked on the website. You can also open it using Rockbox Utility following this strange "Manual" tab
00:20:11Mouser_XHmmm. I should try out this new-fangled RButility...
00:20:14Mouser_X:P
00:20:32Mouser_X(No such thing existed when I got Rockbox.)
00:21:10IsolinearSame here.. :)
00:21:32bluebrothersame for me ... but you should really try it ;-)
00:21:54bluebrother(and testers are always helpful −− I have the impression that quite some users don't report glitches)
00:22:09 Quit davina (Remote closed the connection)
00:22:24IsolinearI got Rockbox before ipodpatcher just wanted you to pick one of three letters for input...
00:23:07bluebrotherI got Rockbox even before it ran on Ipods ... wow, it's quite a while since.
00:23:18Mouser_XWow indeed.
00:23:19 Quit lee-qid ("aufwiederbyebientotsayonara")
00:23:50IsolinearYeah, my first mp3 player was an Archos studio, but I was always afraid to Rockbox it.
00:23:50amiconnBuschel: Hmm, both stride and count can reside on the stack within the loop
00:23:52amiconns
00:24:00amiconnSo you can actually free 2 registers
00:25:52amiconnThat'd be easier in an external .S file as a helper function...
00:26:15Buschelamiconn: can you port it to a .S-file?
00:26:44amiconnShouldn't be difficult
00:27:15 Quit Toxicity999 (Read error: 104 (Connection reset by peer))
00:27:57 Join Toxicity999 [0] (n=bryan@unaffiliated/Toxicity999)
00:28:09*amiconn actually downloads the patch
00:28:23amiconnUntil now I was just reading the plain diff
00:30:50 Join mirak [0] (n=mirak@m94.net81-66-75.noos.fr)
00:31:54Buschelamiconn: gotta leave now, need to go to work in 7h :/ if you can convert the change to a .S-file just post it to fs. thanks for your help and suggestions!
00:32:27amiconnMaybe I'll commit it right away with the suggested changes
00:32:29 Quit HellDragon (Client Quit)
00:32:35amiconnI have a G5.5 for testing
00:32:44mirakhey
00:32:46mirakamiconn: hi
00:32:52 Join HellDragon [0] (n=Nocebo@unaffiliated/helldragon)
00:33:00amiconnBuschel: Of course I'll put you into the (C) line
00:33:04mirakamiconn: I am amazed by mpegplayer, I didn't tried since a long time
00:33:14 Nick Arathis|busy is now known as Arathis (n=doerk@p508A5AF5.dip.t-dialin.net)
00:33:26mirakI am surprised of the number of fps reached even with the sound.
00:33:28amiconnmirak: Yeah, it got some asm love since...
00:33:31Buschelamiconn: then please also test to switch off the 14ms delay (the whole "if finishup_needed"-sequence) and give it a try. works fine during video playback
00:33:47mirakamiconn: I guess not only idct was put in asm
00:34:01amiconnNo, also some variants of the motion compensation
00:34:06mirakamiconn: idct.c I mean. are quantization parts optimised ?
00:34:08mirakok
00:34:11amiconn(the same which are asm'ized for arm too)
00:34:11Buschelbye
00:34:14 Quit Buschel ()
00:35:16mirakamiconn: hum do you think that with wav sound instead of mp3 it could play 25fps without a problem ?
00:35:37preglowaudio is decoded by the second core
00:35:38mirakI would be interested by this possibility
00:35:52mirakpreglow: a second coldfire core ? :-D
00:35:52amiconnI will try to squeeze the y interpolation motion compensation stuff into asm too
00:35:53miraklol
00:35:57preglowmirak: right
00:36:11mirakpreglow: ??
00:36:19preglowthen you will get better fps, yes
00:36:20 Part S1gn
00:36:38amiconnMight be tricky, as I will hit the d register limit - and a registers can't do all the neat things...
00:37:27mirakamiconn: mike something suggested that in idct it could be possible to optimise the byte overflow thing
00:37:36amiconnI also think that there might be some buffers hiding which should be moved into iram
00:38:31mirakamiconn: I looked at asm code for idct but don't remember exactly everything. did you used the double buffer for the 8x8 64 byte pixel block ?
00:38:48amiconnyes
00:38:58amiconnWithout that, the reload would be a pita
00:39:31mirakyep. I did some weird thing also, to be able to store the bytes in reverse order
00:39:32amiconnJust compare your patch with svn :)
00:40:02mirakyes I should. I am happy you finally worked on it :D
00:40:09amiconnI found why your try at an .S file didn't work, and I also applied byte packing to the _add variant of the idct
00:40:51mirakthe .S file was not of me, I didn't had enough knowledge to do that. Maybe you should credit the guy that put the inline into the file, I don't know
00:40:52amiconnThere are some clever tricks hiding, involving swap, shifts and temp registers...
00:41:18mirakamiconn: that you added or that I did ?
00:41:33amiconnme
00:41:35mirakI remember I did that though, but wasn't exactly sure if it was the good way to do it
00:42:52amiconnI also shaved off 2 instructions from the second mac.w block of each loop, utilizing the fact that there are 2 pairs of macs where 2 terms are identical
00:43:17amiconn(using move.l %accM, %accN to duplicate)
00:44:50mirak(MikeS) - Tuesday, 18 September 2007, 00:04 GMT
00:44:50mirakOk, somehow I've gotten into doing this for ARM. Will probably stick with the butterfly form there but initial results with changing the simpler routines looks promising.
00:44:50mirakThere's much room for improvement in this patch since Coldfire can more efficiently 1) use emac multiplication to clamp outputs than it can use shifting. Better yet, 2) scale the emac routines to saturate themselves with no clamping stage by making all output left-justified. Coldfire core DSP uses 1), SPC codec uses 1) and 2). Another word of advice: avoid msac - it's dog slow - and mac the negative product.
00:45:15amiconnI know - and that won't work here
00:45:28mirakah too bad.
00:46:03mirakthis part is a bit furstrating, because it's like there could be an easy optimisation because it's just a poor byte, but in fact no.
00:46:10amiconn..since you need to clamp the final output *after* adding/subtracting the 2 halves of each direction
00:46:37amiconn...but we don't have 8 accumulators, there are just 4
00:46:44mirakamiconn: have you tried implementing the optimised C version ?
00:47:30mirakthe on with 7 mult and 8 adds or something like that.
00:47:42mirakand butterflys
00:48:14mirakIt's not very well suited for EMAC but I am wondering
00:48:38 Quit bluebrother ("leaving")
00:49:47*preglow pleasures himself removing some mallocs and reaping huge performance gains
00:50:04amiconnOh, from where? :)
00:50:08 Quit DogBoy ("Leaving")
00:50:16amiconnmirak: No I didn't
00:50:24mirakamiconn: I was also wondering if at some places, the algorithms works directly on ram, and if it would not be better to try to copy to Iram the subsets of data used, then work on them, and copy back the buffer to ram
00:50:30 Quit ender` (" Sex is dirty - if you do it right. -- Isaac Asimov")
00:50:52n1spixelma: could you test if that midi that gave you a data abort is fixed now with latest svn?
00:50:54mirakfor motion compensation this happens. it works on blocks of 32x32 pixels I think
00:50:56safetydanpreglow, speex?
00:50:56Mouser_Xpreglow: That sounds wrong...
00:51:21preglowsafetydan: yep
00:51:51preglowsafetydan: the decoder states are really small, i get both narrowband and wideband states in iram, no trouble at all
00:51:52safetydanah at last speex is getting some love
00:52:02preglowsafetydan: yeah, i have some assembler lined up too
00:52:14mirakamiconn: from what I remember the only place where there was an alternate buffer was for the idct block. But maybe it could be done for some other stages
00:52:36amiconnmirak: Motion compensation now loads longwords at once instead of bytes
00:52:47pixelman1s: will do in a moment
00:52:53preglowit's a bitch that speex uses 16 bit arrays internally, though, that means dsp has to convert, but again, that's a reason for speex states being so small
00:53:03amiconn(well, the asm optimized variants, which are plain copying and x averaging copying)
00:53:05n1spixelma: thanks
00:53:44 Quit spiorf (Remote closed the connection)
00:54:06amiconnIn fact the asm motion compensation functions load data using movem, so those working on 16x16 blocks might be luckky and even see a line burst occasionally
00:54:10mirakpreglow: would it be hard to allow mpegplayer to use another audio codec ? like wav wavpack or something that really don't use a lot of ressources
00:54:11mirak?
00:54:29amiconnwavpack is more demanding than mpeg audio on coldfire
00:54:40markunI believe ac3 is not so demanding
00:54:45amiconnlibmad on coldfire is very efficient
00:54:59mirakmarkun: gigabeat fares well on the benchs
00:55:24amiconnI think the only compressing codecs with are faster are flac and shorten
00:55:52mirakamiconn: do you have an idea what is the repartition of % cpu usage between libmpeg2 and libmad during playback ?
00:56:04amiconnBut as long as we use mpeg streams we're bound to use mpeg audio - and imho using something else doesn't make much sense
00:56:06preglowmirak: don't know, linuxstb did that
00:56:16preglowbut really, though
00:56:21mirakpreglow: yes right sorry
00:56:25preglowwe should just optimize mpeg audio further in any case
00:56:32preglowit's already pretty fast
00:57:07 Quit Arathis ("Bye, bye")
00:57:12Nico_Ppreglow: I was toying with the spectrum analyser plugin idea yesterday and had a look at a bit of code... does pcm_play_dma_get_peak_buffer provide the required data ?
00:57:24preglowNico_P: no idea
00:57:26mirakamiconn: well wav allows the 25fps on iriver. But that would huge files ...
00:57:44Nico_Ppreglow: do you know how much data is required ?
00:57:57amiconnI'd rather like to see multiple sample rate support in mpegplayer
00:58:14amiconnThen we could use 22kHz audio for less audio decoding demand
00:58:52mirakgood sound is what makes you enter in the video since image is small
00:59:03 Nick _pill is now known as pill (i=pill@sloth.shellfx.net)
00:59:08preglowNico_P: depends how you're going to get the spectrum data, if you're using filters: everything from 1 sample and up.if you're using ffts, it depends on fft block length, usually around 512 or 1024
00:59:25mirakgood night all
00:59:35Nico_Pbytes? how much is asample?
01:00
01:00:03preglowi love the fact that speex does dynamic allocation of data on the stack as much as it can
01:01:12preglowfrom the map file, it looks like current speex only consumes 15kb of iram :P
01:01:19preglownice little codec, this
01:02:54safetydanpreglow, maybe size was more of a problem than speed?
01:02:57 Quit kugel (SendQ exceeded)
01:03:44 Quit jhMikeS (Nick collision from services.)
01:03:47Nico_Ppreglow: what's fastest/easiest between FFT and filters? and the questions above too
01:03:50 Join jhMikeS [0] (n=jethead7@rockbox/developer/jhMikeS)
01:04:08preglowNico_P: fft is fastest if you want plenty of bands, and if you code a good fft
01:04:22preglowfft is probably going to be fastest anyway
01:04:24 Join barrywardell [0] (n=barrywar@host-194-46-226-143.dsl-ie.utvinternet.net)
01:04:33preglowalmost all spectrum analyzers use ffts
01:04:50n1spreglow: you could play with turning up the gcc optimization for speex too, currently it's built with -O
01:05:06Nico_Pok. then does the block size of 512 or 1024 mean several samples are needed ?
01:05:07preglown1s: will do
01:05:34*n1s goes ZzZzZ
01:05:38 Quit n1s ()
01:06:39preglowrealtime ratio is 377% for a wideband (16khz) 19kbps file
01:07:05pixelmajust when I wanted to give n1s the test report...
01:07:29 Nick fxb is now known as fxb__ (n=felixbru@h1252615.stratoserver.net)
01:07:51preglowholy shit
01:07:57preglow427.79% realtime with -O2
01:08:12preglowthis is h120
01:09:11Nico_Phmm apparently good FFT implementations are easy to find, even ASM ones
01:09:20safetydanpreglow, bloody hell that's fast
01:10:01preglowNico_P: they're very common, yes
01:10:03rasherpreglow: how much is that compared to mpeg?
01:10:06preglowsafetydan: that's not even using any asm
01:10:39preglowrasher: can't remember what mpeg is at about now, plus i don't have any 16khz mpeg files
01:10:45rasherAh right
01:11:17rasherJust wondering if speex for voice files might be interesting?
01:11:38preglowit probably will be
01:11:46preglowchecking what realtime ratio is at for pp now
01:11:50preglowprobably less than coldfire
01:12:13 Join saratoga [0] (i=9803c6dd@gateway/web/cgi-irc/labb.contactor.se/x-5d0d09b37f2bb08b)
01:12:23preglowthis is also just one channel, so it's not as impressive as it might sound
01:12:25rasherAh, looks like mpeg is around the same speed for most bitrates on CodecPerformanceComparison
01:12:58safetydanrasher, speex should sound better for smaller files than MP3 though
01:13:03 Join DogBoy [0] (n=john@66-101-59-100-static.dsl.oplink.net)
01:13:04preglowpp gets 284%, yes
01:13:34rashersafetydan: yeah, should be the right choice, if the decoding speed gets useful
01:14:31***Saving seen data "./dancer.seen"
01:15:15safetydanpreglow, I'll be interested to see what changes you made. I had a go at iramifying things a while back but had no idea what I was doing so it didn't amount to much.
01:16:05preglowsafetydan: commit coming up in five minutes
01:16:20 Quit roolku ()
01:16:21preglowdoing some more testing
01:16:47 Join HSorgYves [0] (i=HSorgYve@unaffiliated/pi/x-000000001)
01:16:55 Part HSorgYves
01:17:28preglowgeh, arm is slower with O2
01:17:47 Join JdGordon [0] (n=jonno@c210-49-113-143.smelb1.vic.optusnet.com.au)
01:19:00preglowi'll just hack that into the makefile, then
01:19:43feindbildhi =)
01:21:40feindbildI have some hints for encoding video with mencoder. Where should I put them so others don't have to spend weeks figuring it out? the wiki?
01:22:04 Join kugel|afk [0] (i=kugel@unaffiliated/kugel)
01:22:10 Nick kugel|afk is now known as kugel (i=kugel@unaffiliated/kugel)
01:22:26 Quit jhulst (Remote closed the connection)
01:24:06preglowsafetydan: there
01:24:13pixelmafeindbild: probably in the PluginMpegplayer wiki page as it already contains info about encoding (also with mencoder)
01:24:54preglowalso, please test speex files now, i don't have many test files
01:25:36 Quit mirak (Remote closed the connection)
01:26:23safetydanpreglow, I'll take a look when I have my h120 to hand again. I have a directory of test files to go through
01:26:52preglowsafetydan: excellent
01:28:31preglowwtf, my nano just idled, and now it has prefetch abort at 0xdeadbeee
01:28:59preglowamiconn: i told the speex author about the static with no inline problem and it's fixed in latest svn
01:29:05 Quit barrywardell ()
01:29:11scorchedid the c0edbabe swat it?
01:29:36feindbildpixelma: hmmm ... somebody interested in putting it up if I post it to a pastebin? ^^
01:30:18scorchefeindbild: well, it is a wiki..
01:30:42safetydanpreglow, you didn't have a chance to look at stereo decoding did you?
01:31:10preglowsafetydan: i have, and it looks like it'll be simple to at least have the inner loop fixed point-ized
01:31:18preglowit's just a simple iir filter weighting each channel
01:31:26preglowbut it also contains some libm calls
01:31:37feindbildscorche: well ... one that only allows using your realname as wikiname
01:31:38*feindbild ducks
01:32:07scorchefeindbild: yes?...why would that be a hindrance?
01:32:26preglowsafetydan: hmm, only two sqrts, though, should be simple
01:32:59feindbildscorche: sorry, I'm kind of paranoid =P never mind ...
01:33:54scorchefeindbild: well, it is just a name...attributed to technical works (or other sorts)...it isnt like you are giving your address, phone number, etc
01:35:09feindbildscorche: hrhr ... except my name is uniqe and thus I'm giving that information away^^ like I said ... never mind.
01:36:10krazykitfeindbild, better not introduce yourself to anyone then ;-)
01:36:20scorcheand what would a malicious person do with that name?..."Xxxx Xxxx put something on rockbox's wiki!...this is a horrible act!"
01:37:37 Join jeremyb [0] (n=jeremy@unaffiliated/jeremyb)
01:38:08kugelfeindbild: I dount your name is as unique as you think
01:38:46Mouser_XI searched for my name on Google. Apparently, I'm a script writer.
01:40:05kugelWhen I search for my name on google it shous only tracker entries^^
01:40:11kugelshows*
01:40:31*kugel stops being off topic
01:41:55safetydanpreglow, darn, doesn't look like much of those changes can be contributed back to speex
01:43:46makuserucan someone please tell me how to organize the music on a player with rockbox, i cant find anywhere where it says how
01:44:08 Quit Thundercloud (Read error: 104 (Connection reset by peer))
01:44:12Mouser_Xmakuseru: Use the database.
01:44:41feindbildn8 =)
01:44:43 Part feindbild ("http://www.last.fm/user/_poison/")
01:45:00 Part pixelma
01:45:20makuseruMouser_X: when i click on Database it says "Database is not ready Initalize now SELECT = Yes Any Other = No" and i cant figure out which button is select, ive pressed any buitton and nothing happened
01:46:25Mouser_Xmakuseru: If you don't know which button is select, then you didn't read the manual...
01:47:01preglowsafetydan: i wouldn't expect them to be
01:47:11makuserui just looked all around the site that had to do with my player, i didnt want to go through 156 pages
01:47:17Mouser_Xhttp://download.rockbox.org/manual/rockbox-gigabeatf/rockbox-buildch4.html#x7-380004.2
01:47:17preglowsafetydan: the asm will be, though
01:47:22Mouser_X^ Database
01:47:37Mouser_X( @ makuseru)
01:47:59makuseruthank you very much
01:47:59Mouser_XAlso @ makuseru: http://download.rockbox.org/manual/rockbox-gigabeatf/rockbox-buildch3.html#x5-210003
01:48:36 Quit Sedgewick (Read error: 110 (Connection timed out))
01:48:47markunmakuseru: and you don't need to read all of the manual of course, just look for the sections you are having questions about
01:49:24*preglow wonders if profiling still works
01:58:16kkurbjunwhat endianness do we expect the arm targets to be running as?
02:00
02:00:41*amiconn wonders whether this asm code will work in the 1st try :>
02:03:06preglowmy experiences with the matter indicates an almost certain "no" :D
02:05:19 Quit japc (Remote closed the connection)
02:05:51 Quit animeloe ("Leaving")
02:09:45 Join animeloe [0] (n=animeloe@unaffiliated/animeloe)
02:15:29 Quit jhMikeS (Nick collision from services.)
02:15:35 Join jhMikeS [0] (n=jethead7@rockbox/developer/jhMikeS)
02:15:43amiconnyay! :D
02:16:41amiconn19fps with the widescreen version of elephants dream on ipod video isn't bad, is it?
02:17:00amiconnOkay, it varies somewhat, now 1.5
02:17:04amiconn*17.5
02:18:53amiconnFullscreen runs at ~13fps
02:20:19preglownot bad at all!
02:22:53saratogain 64 bit linux, do i have to import something special to get int32_t types?
02:24:00Soap13 fps is only 1 more than the wiki claims.
02:24:19Soapand IIRC that number (12) was before the recent optimizations.
02:25:35safetydansaratoga, #include <inttypes.h> or something like that?
02:28:06*preglow wonders why his h120 sometimes boots into reatilos after using the reset button
02:28:43preglownot even sometimes, it does so quite reliably
02:29:14amiconnSoap: The numbers in the wiki are imho a bit too optimistic if they are really from be fore the C optimisation
02:29:44amiconnMaybe they're taken using just the very beginning of elephants dream, which seems to decode easier than later parts
02:30:07amiconn160x128 is now running at 37fps
02:30:27preglowsafetydan: have you ever tried decoding ultra-wideband mode files?
02:32:14Soapamiconn: they are - I did them
02:34:08Soap(they are from before the optimization) - someone tweaked the last two - but the methodology I employed was to watch the entire movie through to the end - write down the ending FPS, then watch the movie again and find two places early on in the film with a matching FPS, and use those places as my "benchmark" for all future tests.
02:34:09 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
02:34:48SoapI ASSumed this would give me decent numbers w/o having to watch the film repeatedly @ sub 20fps.
02:38:07 Quit Nico_P (Remote closed the connection)
02:41:21amiconnpreglow: Hmm, I hope your latest commit doesn't cause problems with voice?
02:41:41safetydanpreglow, I think so and they should work. UWB is for 16 khz and higher isn't it?
02:41:57preglowsafetydan: 32khz
02:42:01preglowamiconn: why should it?
02:42:02amiconnNm, didn't look close enough
02:42:07preglowcan do
02:43:11preglowsafetydan: uwb files hang here, don't know if that's my fault
02:44:27preglowmight also be that they use float and just seem to hang
02:47:01preglowsegfault in sim, hrm
02:49:51safetydanpreglow, I haven't actually tested them since the original commit for speex months ago
02:50:11preglowme neither, i thought i had a uwb file, but i see now that it is 44.1khz wb
02:50:16preglowprobably why it sounds so bloody awful
02:52:09 Quit tchan (Read error: 104 (Connection reset by peer))
02:52:48 Quit saratoga ("CGI:IRC")
02:53:52 Join tchan [0] (n=tchan@lunar-linux/developer/tchan)
02:56:12preglowthink i'll check this out tomorrow
02:56:14preglowbedtime now, nightie
02:56:57 Join barrywardell [0] (n=barrywar@host-194-46-226-143.dsl-ie.utvinternet.net)
02:57:52amiconnCombining Buschel's and my own observations, I suspect that the broadcom "finishup" is in fact not a necessary finishup function, but rather a frame sync feature, that syncs transfer to the internal frame frequency
02:58:38amiconnIt also seems that the broadcom has some kind of internal anti-glitch measure that delays writes if they come too slow
02:58:39 Quit jhulst (Read error: 110 (Connection timed out))
02:59:31amiconnSo it might in fact be fastest to render into an intermediate ram buffer first, then transfer than. And lcd_update() can be optimised more than what that duff's device compiles into
02:59:44amiconns/than/that/
03:00
03:00:12Soapamiconn: do you think my FPS measuring methodology is flawed?
03:00:51amiconnMeasuring for a too short time is bad, as well as stopping at a position where fps tend to vary
03:01:29amiconnOtherwise it should be fine, although measuring the whole video would be more precise
03:01:54amiconnIf my theory is correct, that should open the door for significant fps gains...
03:02:11Soapyour Broadcomm theory?
03:02:15amiconnyup
03:02:48SoapIf your Broadcom theory is correct, then why has the Video long had framerates similar to the Photo (w/o a Broadcom chip)?
03:03:08amiconn?
03:06:33amiconnAnyway, it's a theory that's worth checking. But not right now
03:06:47Soapsince the processor is (basicly) the same between the Photo and the Video - why (for example) does the photo get 25 fps @ 224x176, and the Video gets 26 (identical within the margin of measuring error)
03:07:20amiconnYOu can't compare those. It's a totally different hookup. ANd the Photo will get much better frame rates soon
03:07:22SoapI would assume if the video was getting bottlenecked by the Broadcom, it would have lower FPS numbers than the Photo @ common resolutions.
03:08:00*amiconn goes to sleep
03:08:05SoapOK, I can accept that - why I asked.
03:11:04 Quit barrywardell ()
03:13:27 Quit Soap (Read error: 104 (Connection reset by peer))
03:14:32***Saving seen data "./dancer.seen"
03:15:43 Join amigan_ [0] (i=dcp1990@ip70-181-22-88.ri.ri.cox.net)
03:16:45 Join AndrewJ [0] (n=KeyLime@70.232.140.147)
03:16:48rasherkkurbjun: m:robe sim still seems to fail
03:18:30JdGordonrasher: fail how? compiling?
03:18:45rasherYeah
03:18:54JdGordoncompiles fine here.. as long as you only do make bin
03:18:59rasherOr perhaps just when crosscompiling
03:19:17rasherOh, make install fails?
03:19:19JdGordonhaha.. the sim takes up half my deksopt!
03:20:27 Join PaulJam [0] (i=PaulJam_@vpn-3002.gwdg.de)
03:21:34rasherNope. Still doesn't build: pluginlib_actions.c:125:6: error: #error pluginlib_actions: Unsupported keypad
03:22:06 Quit kugel ("Benutzer ist abwesend.")
03:22:38JdGordonok, yeah.. fixing that now
03:22:54AndrewJ?
03:23:09 Part AndrewJ
03:23:13 Quit PaulJam (Client Quit)
03:24:05 Join emeraldd [0] (n=jules@h96.156.30.69.ip.alltel.net)
03:26:05 Join Soap [0] (n=Soap@rockbox/staff/soap)
03:29:24JdGordonrasher: fixed, the sim comiles now
03:32:34 Quit amigan_ (Connection timed out)
03:34:46rasherJdGordon: it does indeed. Crikey it's huge. Something's acting up with −−background though
03:34:59JdGordonyeah, saw that
03:35:04JdGordonit needs a proper image
03:35:15*JdGordon wonders what −−zoom would look like!
03:35:35JdGordon480x640 is a bit huge :p
03:36:41 Quit amigan (Read error: 110 (Connection timed out))
03:37:02rasherEven ter-32b isn't terribly large
03:44:16 Join psycho_maniac [0] (i=psycho_m@ppp-64-91-85-210.cam.centurytel.net)
03:44:21 Quit makuseru (Read error: 104 (Connection reset by peer))
03:50:22advcomp2019i have an issue with r15256 on my sansa e280r.. it takes like two minutes to start playing a song when i hit resume playback from when i turn it on
03:55:25emeralddwould it be possible for someone to set the svn:ignore property to ignore build directories?
04:00
04:02:36advcomp2019odd it is not doing it now
04:02:55safetydanemeraldd, which ones though? I have about 6 different build directories
04:03:05safetydanit should at least ignore build/ at the moment
04:03:58 Join mrkiko [0] (n=mrkiko@adsl-ull-185-119.42-151.net24.it)
04:04:52emeralddIf I do an svn status, I get a ? build
04:04:56emeralddfor a build/
04:05:20mrkikoHi all! Someone knowing something about FS #8003? Has been some progress made? Anyone is free to contact me for testing purposes!
04:05:21emeralddalthough it doesn't show up in svn diff
04:19:38 Quit Gnu47 (Nick collision from services.)
04:19:45 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
04:19:57 Quit donutman25 ("ChatZilla 0.9.78.1 [Firefox 2.0.0.8/2007100816]")
04:25:55 Quit Gnu47 (Nick collision from services.)
04:26:03 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
04:34:38 Join karashata [0] (n=Kimi@pool2-086.adsl.user.start.ca)
04:36:13 Quit Gnu47 (Remote closed the connection)
04:36:25 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
04:42:06 Quit hcs (Read error: 104 (Connection reset by peer))
04:42:32 Join hcs [0] (n=agashlin@rockbox/contributor/hcs)
04:48:50 Part mrkiko ("rockbox")
04:51:55 Nick EnterUse1Name is now known as EnterUserName (n=dave@ip-128.55.99.216.dsl-cust.ca.inter.net)
04:55:14 Quit Gnu47 (Read error: 110 (Connection timed out))
04:58:15 Join mrkiko [0] (n=mrkiko@adsl-ull-185-119.42-151.net24.it)
05:00
05:01:27 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
05:09:25mrkikorasher: regarding the problems on h300 series readers (crash): you're right. It may crash occasionally even not interrupting the voice. Aniway it happened for now only when the voice features are activated.
05:14:34***Saving seen data "./dancer.seen"
05:16:05 Quit bb (Nick collision from services.)
05:16:10 Join bb_ [0] (n=bb@unaffiliated/bb)
05:16:25 Join z35 [0] (n=z@149.123.33.65.cfl.res.rr.com)
05:22:19 Quit emeraldd (Remote closed the connection)
05:22:36 Join emeraldd [0] (n=jules@h96.156.30.69.ip.alltel.net)
05:27:52 Join iamben [0] (n=ben@dpc67142179038.direcpc.com)
05:28:51 Quit hannesd (Read error: 110 (Connection timed out))
05:29:44 Join FM [0] (n=fortemas@ottawa-hs-206-191-39-130.d-ip.magma.ca)
05:31:45FMI have a stupid question to ask, and I did RFTM: I booted Rockbox. I turned off the player and now it won't seem to go back on.
05:31:59FMI have a 5G iPod.
05:32:55FM...huh. It just went back on.
05:34:43 Quit nerochiaro (Remote closed the connection)
05:35:43 Quit male ("User disconnected")
05:38:07scorcheglad we could help! :)
05:45:27emeralddis something up with the build process? http://build.rockbox.org/dev.cgi is claiming that a build should have happened som 779 minutes ago
05:46:25psycho_maniacwhere you see that?
05:46:29 Quit hcs ("Leaving.")
05:47:43psycho_maniacoh ok i see it.
05:48:30safetydansomeone in a european time zone will need to poke it
05:49:22scorcheswedes
05:50:49JdGordonoh bugger
05:51:19JdGordonhopefully zagor comes online today then
05:52:31emeraldddoes it blow up like that very often?
05:52:38scorcheit hangs at times
05:52:44JdGordonnot often, but enough to be annoying
06:00
06:01:25safetydan<off_topic>I always laugh when someone says 'swede' since my first thought is the vegetable</off_topic>
06:02:49 Quit criznach ("ChatZilla 0.9.78.1 [Firefox 2.0.0.8/2007100816]")
06:03:54scorcheo_O
06:04:05*scorche never heard of any swede vegetable
06:05:20*alienbiker99 never heard of it either
06:05:38JdGordonjhMikeS: you round?
06:05:59safetydanFrom wikipedia: # Rutabaga, a root vegetable, commonly called "swede".
06:06:11scorcheah...didnt know that
06:06:12safetydanmay be an commonwealth thing
06:06:34*JdGordon never heard of the vegie swede
06:06:39JdGordonbut im not a vegie man anyway
06:07:14 Quit atsea-22 (Read error: 104 (Connection reset by peer))
06:09:23jhMikeSJdGordon: yeah
06:09:35JdGordondo you tinhk http://pastebin.com/m7a5629e4 is ok?
06:09:50JdGordonfor getting button data out fo the drivers
06:10:21 Quit Mouser_X (Read error: 110 (Connection timed out))
06:10:22JdGordonits adds a data variable which im hoping gcc will replace with 0 in all but the mrobe targets
06:10:58jhMikeSJdGordon: I notices a strange behavior that shows up in the 10/20 build. On my e200, if I start a database update and try to browse files, it will often hang until the database refresh is complete. Don't know if the list changes could do something.
06:11:41jhMikeSAuto-Update = off (BTW)
06:12:15JdGordonhmm... seems to happen here also :(
06:12:20CalcipherjhMikeS you have a e200 series aswell?
06:12:46Calcipherseems like a few folks here do
06:13:05JdGordonjhMikeS: i dont know why that list changes would affect it.. maybe Slasheri has an idea
06:13:18 Join Mouser_X [0] (n=someone@207.155.176.3)
06:13:38jhMikeSCalcipher: yep
06:13:59Calciphercool what model did you get?
06:14:12jhMikeSJdGordon: That looks good at first glance.
06:14:17jhMikeSCalcipher: e260
06:14:47CalcipherI was wondering, if tehre was any difference in the hardware between the e200 and the e200r models, or was it just software?
06:14:50 Quit z35 (Read error: 104 (Connection reset by peer))
06:15:03 Quit RoC_MasterMind ("Leaving")
06:15:04scorchesoftware
06:15:29Calcipherah good, guess thats why the patch to use the e200 bootloader works so well
06:15:34scorchethere is the matter of the fm tuner, but some e200s have them, some dont
06:15:53jhMikeSJdGordon: why bother making button data target specific?
06:16:36Calcipherah ok
06:17:14 Quit emeraldd ("using sirc version 2.211+KSIRC/1.3.12")
06:17:44jhMikeSJdGordon: I thought you were posting button data directly for m:robe like a wheel driver does
06:17:58 Join z35 [0] (n=z@149.123.33.65.cfl.res.rr.com)
06:18:59JdGordoni was, but that sucked because i didnt want to redo the release/repeat handling manually
06:19:23JdGordonand its target specific because all but 3 targets actually have some data to pass
06:20:34 Quit ToHellWithGA ("You know you'll miss me a lot.")
06:20:35jhMikeSyeah, you're right, it makes sense to do
06:21:11 Join Leif_Erikson [0] (n=chatzill@75.139.212.31)
06:22:00JdGordonno bin difference on the ondio build, so ill put it in
06:23:05jhMikeSno increase using data anyway?
06:23:17 Quit lazka (Remote closed the connection)
06:23:18JdGordon-4 bytes even
06:23:24jhMikeSheh
06:23:37JdGordonbbl
06:25:51 Quit mrkiko (Remote closed the connection)
06:26:58 Quit FM ("The light to the darkness; the darkness to the light.")
06:38:07 Quit Mouser_X (Nick collision from services.)
06:38:27 Join SkinInd95 [0] (n=chatzill@host-69-144-93-208.hln-mt.client.bresnan.net)
06:39:05 Join Mouser_X [0] (n=someone@207.155.176.3)
06:44:54 Quit luckz (Remote closed the connection)
06:44:55 Join makuseru [0] (n=max@163.106.40.24.aeneasdsl.com)
06:46:39 Quit SkinInd95 ("ChatZilla 0.9.78.1 [Firefox 2.0.0.8/2007100816]")
06:46:53makuseruhow can i create my own wps?
06:49:18karashatahttp://www.rockbox.org/twiki/bin/view/Main/CustomWPS <−− read that
06:50:13makuseruthank you
06:50:20karashatayou're welcome
06:56:51Leif_EriksonI'm trying to get Rockboy to work on my 5g ipod, I selected the .gbc file from the file viewer and nothing happened. Is there something I missed on the website?
06:56:55BHSPitMonkeydaurnimator, yo
06:59:01 Join luckz [0] (n=luckz@luckz.de)
06:59:04daurnimatorBHSPitMonkey?
06:59:24BHSPitMonkeydaurnimator, you still roam toc2rta?
06:59:40daurnimatoruh, I went in there 2 days ago for about an hour
07:00
07:00:26BHSPitMonkeyjust trying to figure out why I can't connect
07:01:00daurnimatori just connected
07:01:10daurnimatornote: no more psp channels there - just iphone
07:01:33Mouser_XDang.
07:01:35BHSPitMonkeyyeah, I know- I've been in iTouch for a couple days at home
07:01:42BHSPitMonkeywow
07:01:46 Join qwm [0] (n=qwm@h38n2fls32o1010.telia.com)
07:01:53BHSPitMonkeyit worked in irssi, but not in xchat. I wonder what the heck could be causing it.
07:02:43 Quit XavierGr (Nick collision from services.)
07:02:46 Join XavierGr [0] (n=xavier@ppp193-119.adsl.forthnet.gr)
07:03:04qwmhm, is time stretching possible with rockbox?
07:03:40BHSPitMonkeyqwm, I actually used my iPod to make a boring lecture go by quicker last week
07:03:54scorcheqwm!
07:04:06daurnimatorI spose if you listen to a really boring audio bookj
07:04:17scorcheqwm: not at the moment
07:04:41karashataif you adjust the pitch, it is
07:04:51scorchethat isnt time-stretching
07:04:55karashataoh
07:04:59karashatanvm then
07:05:00qwmhehe.
07:05:05qwmi want the pitch intact.
07:05:18scorchetime-stretching is with the pitch the same as normal
07:05:23karashatayeah, then it's not possible
07:05:24karashatayet
07:05:31karashatasomeday, who knows?
07:05:32scorchewhich is what i said ;)
07:05:39karashataahh
07:07:06qwmmake it happen, scorche
07:07:07qwm:p
07:07:20*scorche snaps his fingers
07:07:22scorcheit is done
07:07:39Mouser_Xyay!
07:07:46 Join atsea-22 [0] (i=atsea-@gateway/tor/x-00a308ddff3fa4b4)
07:12:05 Quit jhulst ("Konversation terminated!")
07:14:38***Saving seen data "./dancer.seen"
07:19:09 Part Leif_Erikson
07:27:31 Part safetydan
07:36:21 Nick fxb__ is now known as fxb (n=felixbru@h1252615.stratoserver.net)
07:37:10 Join criznach [0] (n=chatzill@host-69-145-134-192.grf-mt.client.bresnan.net)
07:44:20 Nick fxb is now known as fxb__ (n=felixbru@h1252615.stratoserver.net)
07:45:22 Join homielowe [0] (n=chatzill@d207-81-67-190.bchsia.telus.net)
07:47:27 Quit Mouser_X (Read error: 104 (Connection reset by peer))
07:48:26 Quit qweru ("moo")
07:55:12 Join Mouser_X [0] (n=someone@207.155.176.3)
07:56:25makuseruif i wanted to just changed a background on a theme could i not just replace the image that the config file points to?
07:57:08karashatadepends, some of what's displayed might be part of the background image
07:57:51karashataand that would only change the background that's used in the menus anyway, you'd have to replace the background file (whatever it's name is) in the WPS folder to change the WPS background
07:57:52Mouser_Xmakuseru: As long as it was the right resolution. Of course, as karashata said, this will only work if it's the backdrop, and not part of the WPS screen.
07:58:22makuseruim in the backdrop folder
07:58:39makuseruoooooh
07:58:46makuserui was wondering why it wouldnt work
07:58:58makuserui had the size mixed um
07:58:59makuseruup*
08:00
08:07:03 Part toffe82
08:15:16 Quit psycho_maniac (" HydraIRC -> http://www.hydrairc.com <- The alternative IRC client")
08:15:20 Join ddalton [0] (n=Daniel@203-214-76-97.dyn.iinet.net.au)
08:15:46ddaltonJdGordon: what code is executed when the user presses up or down in the file browser?
08:16:02 Quit lids ("leaving")
08:17:23 Join advcomp2019_ [0] (n=advcomp2@66.172.231.192)
08:17:37 Quit advcomp2019 (Nick collision from services.)
08:17:43 Nick advcomp2019_ is now known as advcomp2019 (n=advcomp2@66.172.231.192)
08:18:34 Join spiorf [0] (n=spiorf@host98-168-dynamic.6-79-r.retail.telecomitalia.it)
08:19:56 Quit Mouser_X (Read error: 110 (Connection timed out))
08:23:27 Quit BigBambi (Read error: 113 (No route to host))
08:26:34 Join davina [0] (n=davina@cpc1-sout6-0-0-cust616.sotn.cable.ntl.com)
08:26:58jhMikeSddalton: I think he's away for a bit
08:27:19ddaltonjhMikeS: ah ok
08:32:20 Join GodEater_ [0] (n=bryan@rockbox/staff/GodEater)
08:39:58 Join Rob2222 [0] (n=Miranda@p54B1671C.dip.t-dialin.net)
08:40:34 Quit XavierGr (Nick collision from services.)
08:40:37 Join XavierGr [0] (n=xavier@ppp176-207.adsl.forthnet.gr)
08:44:19 Quit spiorf (Read error: 110 (Connection timed out))
08:44:30 Quit darkless ("Leaving")
08:44:53*amiconn had an idea how to squeeze quite some cycles out of the libdemac filter math for arm :)
08:45:31amiconnUnfortunately this won't help to get it playable on PP, as there even -c1000 is too slow, but it will help the gigabeat
08:45:50amiconn(-c1000 doesn't use that filtering at all)
08:46:38JdGordonamiconn: do you see any problems with http://pastebin.com/m7a5629e4 ?
08:47:02JdGordonits that or more IF_xx() inline macros :(
08:47:22ddaltonJdGordon: do you know what is executed when moving through the file browser?
08:47:26ddaltonup and down
08:48:09JdGordonddalton: you keep asking the same bloody question... look in the file where the code for the screen is.. then look for the ACTION_STD_ if/switches...
08:48:36 Join ender` [0] (i=krneki@84-255-206-8.static.dsl.t-2.net)
08:48:39 Join miepchen^schlaf [0] (n=hihi@p54BF7726.dip.t-dialin.net)
08:49:31 Quit makuseru (Read error: 104 (Connection reset by peer))
08:50:12amiconnJdGordon: What kind of problems?
08:50:24JdGordonwith how the ifdeffing is done
08:50:47ddaltonok...
08:50:49JdGordonalso, is that extra data varilable going to add to bin size?
08:51:14JdGordoni checked and dont think so, but the build serer is donw now so i figured id get another pair of eyes while we wait
08:51:32amiconnNo its not
08:51:38amiconn...since it's always 0
08:51:53JdGordonk great, thats what i was hoping
08:51:58amiconnYou could give the compiler one more hint that it will stay 0 all the time
08:52:14 Quit karashata ("Leaving.")
08:52:23JdGordonadding const?
08:52:43amiconn#ifdef HAVE_BUTTON_DATA / int data = 0; / #else / const int data = 0; / #endif
08:53:04JdGordonnot pretty, but could save hassle later if somone tries modifing it accidently
08:53:16amiconnyes
08:54:07 Nick bb_ is now known as bb (n=bb@unaffiliated/bb)
08:54:51 Join bertrik [0] (n=Bertrik_@031-020-045-062.dynamic.caiway.nl)
08:56:04ddaltonJdGordon: re my tree question. every button is there except down up left and right.
08:56:16ddaltonthere is rec, menu, long menu, and context.
08:56:40JdGordonup/down is handled in do_synclist_button
08:56:45*ddalton Hasn't had a compile error yet...
08:56:47JdGordonleft is _CANCEL and right is _OK
08:56:56ddaltonoh ok
08:57:13ddaltonis do_synclist_button in tree.c?
08:58:30 Quit Rob222241 (Read error: 110 (Connection timed out))
08:58:50JdGordongui_synclist_do_button() is gui/list.c
08:59:23*JdGordon thinks the touchpad driver is ready for svn :)
09:00
09:00:58 Join merbanan [0] (n=banan@83.233.242.209)
09:09:05 Join Zagor [0] (n=bjorn@rockbox/developer/Zagor)
09:10:04 Quit sup (Read error: 104 (Connection reset by peer))
09:11:08 Join MattBatt [0] (i=482893ed@gateway/web/cgi-irc/labb.contactor.se/x-8ce691e062d413d4)
09:11:36 Quit Isolinear (Read error: 110 (Connection timed out))
09:12:38 Join sup [0] (i=1000@80.216.235.247)
09:13:23 Quit miepchen^schlaf (Read error: 110 (Connection timed out))
09:14:03MattBattHi I'm MattBatt I was told by this page http://www.rockbox.org/twiki/bin//view/TWiki/TWikiRegistration to come here and ask for write permissions on the Twiki Thanks
09:14:39***Saving seen data "./dancer.seen"
09:14:56 Part MattBatt
09:15:39ddaltonJdGordon: can you tell me what I am not voicing in the patch I just uploaded? FS #8012. it is about the info list.
09:15:43ddaltonI can't work it out...
09:19:33 Join petur [0] (n=petur@rockbox/developer/petur)
09:19:36*ddalton Is wondering why JdGordon won't talk to him
09:20:53 Quit bertrik ("bye")
09:28:07 Join LinusN [0] (i=linus@rockbox/developer/LinusN)
09:29:34ddaltonLinusN: what do you think Of FS #8007 and FS #8006?
09:30:18LinusNddalton: i just came back from a week in china, so i haven't looked at flyspray yet
09:32:43ddaltonLinusN: ah ok
09:32:45 Join pixelma [0] (i=pixelma@rockbox/staff/pixelma)
09:36:23 Join pondlife [0] (n=Steve@rockbox/developer/pondlife)
09:36:52 Join webguest52 [0] (i=dcef09c8@gateway/web/cgi-irc/labb.contactor.se/x-c8e802fa4fea9641)
09:38:07 Quit webguest52 (Client Quit)
09:40:45 Quit TMM (Read error: 110 (Connection timed out))
09:41:30 Quit ddalton ("leaving")
09:42:56 Quit iamben (Read error: 104 (Connection reset by peer))
09:46:08 Join ddalton [0] (n=Daniel@203-214-76-97.dyn.iinet.net.au)
09:47:30 Join iamben [0] (n=ben@dpc67142179038.direcpc.com)
09:47:39JdGordonwelcome back LinusN, ho was the trip?
09:48:01LinusNwonderful!
09:48:19JdGordonmind kicking the build server please?
09:48:36 Join B4gder [0] (n=daniel@rockbox/developer/bagder)
09:50:37ddaltoncan all players play a beep?
09:51:57 Join TotallyInfected [0] (n=ebola@pool-70-110-242-209.phil.east.verizon.net)
09:52:14 Join Nico_P [0] (n=nicolas@rockbox/developer/NicoP)
09:52:38ddaltonCan someone give me an url to get a build for h300 of rev 15226?
09:52:49ddaltonim trying to track down a small problem...
09:54:37GodEater_ddalton: can't you just build one yourself ?
09:55:15B4gder15236 and 15216 are available
10:00
10:01:04 Quit criznach ("ChatZilla 0.9.78.1 [Firefox 2.0.0.8/2007100816]")
10:02:08ddaltonGodEater_: I guess I am now but it takes for ever...
10:02:57ddaltonGodEater_: are you having any problem where you lose settings on a shutdown...
10:02:57pondlifeNico_P: Good progress over the weekend it seems. Are you planning to rebase soon?
10:03:10Nico_Ppondlife: probably, yes
10:03:37Nico_Ppondlife: have you tested the latest versions ?
10:03:51pondlifeOK, I might wait until you've done that before testing further - given that I was having crashes with SVN too on Friday.
10:03:59pondlifeNot since Friday
10:04:03Nico_Pok
10:04:06pondlifeBut should have a good play today
10:04:22*Nico_P tries rebasing now
10:04:34pondlifeI've not checked IRC logs yet, but did you and jhMikeS discuss threading much more?
10:04:48*pondlife wants to read and learn
10:05:00Nico_Pnot much more I think
10:05:15pondlifeOK
10:05:25 Join miepchen^schlaf [0] (n=hihi@p54BF44F7.dip.t-dialin.net)
10:05:57pondlifeNico_P: One thing that looks odd to me - do you really need playback.c to have a "buffering" bool ?
10:06:10pondlifeI'd think it should come out in the wash in buffering.c
10:06:15Nico_Pit only reads the one from buffering.c
10:06:28pondlifeDoes that have to be public?
10:06:47pondlifei.e. calllers shouldn't need to know the internal state
10:06:57pondlifeIt should "just work"
10:07:01Nico_Pothere threads can find it useful to know the status... I'll try to find something better though
10:07:16Nico_PI think that atone point I'll have to rethink the audio thread logic
10:07:26Nico_Prebasing is done btw
10:07:33pondlifeCheers, will rebuild
10:07:50pondlifeI'd think that the mod in http://repo.or.cz/w/Rockbox.git?a=commit;h=df2f17234ba1fec7518733b5187a52c5e53d90ca should be inside buffering.c
10:08:14pondlifei.e. so that it just ignores extraneous requests
10:08:45Nico_Pit does ignore them but the audio thread will just keep posting them in a loop, which is a bit of a waste
10:08:51Nico_Pand is annoying in the sim too
10:09:11pondlifeAh, but not actually causing a crash or corruption?
10:09:21Nico_Pno
10:09:23pondlifeSo this is an optimisation, not a fix?
10:09:35Nico_Pbasically, yes
10:10:15JdGordonhey B4gder (welcome back), can you kick the build server please?
10:10:33*Nico_P has to go to class
10:10:41JdGordonno you dont
10:10:46JdGordonclass is bad for you...
10:10:51Nico_Phehe :)
10:10:53*B4gder kicks
10:10:59pondlifeHehe, from holidaying in China to kicking a build server... welcome back to real life :)
10:11:25pixelmait wasn't holidays ;)
10:11:44 Quit Nico_P (Remote closed the connection)
10:12:31 Join Nico_P [0] (n=nicolas@rockbox/developer/NicoP)
10:12:41pondlifeWell, I was trying to emphasize something or other... :)
10:12:59*ddalton Is having great fun debugging rockbox :-)
10:14:50 Join safetydan [0] (n=safetyda@rockbox/developer/safetydan)
10:15:20 Join obo [0] (n=obo@rockbox/developer/obo)
10:18:42*ddalton Can't be bother debugging any more. He will just let someone else complain...
10:21:55pondlifeNico_P: Quick class?
10:22:33pondlifeWhen you get back, your current code isn't rebuffering, it just runs out of data
10:23:09pondlifeIf I stop and restart, it buffers one track
10:25:23*GodEater_ won't bother building that then
10:27:12 Quit merbanan (Remote closed the connection)
10:29:42pondlifeGodEater_: Actually, it only happened once... is now buffering... weird!
10:29:57pondlifeI implore you to reconsider
10:31:42GodEater_perhaps I shall then ;)
10:31:58*GodEater_ just wants to get his FS7738 into the build too.
10:33:50 Join webguest82 [0] (i=50bb6b01@gateway/web/cgi-irc/labb.contactor.se/x-f43f57bac5911da4)
10:34:35 Join Sedgewick [0] (n=10334AEF@host99-233-dynamic.15-87-r.retail.telecomitalia.it)
10:35:23webguest82hi, would it possible to reset the running time to zero automatically each time the players charger is plugged in ?,
10:35:53LinusNwebguest82: it is supposed to do that, afaik
10:36:10webguest82er nope not on my H140
10:39:05webguest82by the way LinusN, have you seen CC's patches 7898 and 7911, ? I like the way he has simplified the playlist Insert menus
10:42:11LinusNwebguest82: remember that it can only reset it if you plug in the charger while rockbox is running
10:42:46webguest82aha, thanks for the info
10:44:11ddaltonLinusN: had a chance to take a look at the tracker?
10:44:58 Part webguest82
10:47:17 Quit ddalton ("no one wants to talk to me")
10:54:44amiconnLinusN: It *is* supposed to do that, but it's broken for quite some time now :/
10:54:55LinusNoops, bad
10:55:05amiconn(some time == months)
10:55:55LinusNhow come?
11:00
11:00:19 Join igorr [0] (n=igorr@oinn.ifi.uio.no)
11:02:30igorrHi, I have just tried rockbox last night on my iRiver-iHP140 and I am really impressed with it. I see that there are some plugins that it does not support, albeit newer players do. One of them is the clock plugin. Does anyone know if there is something in the architecture fundamentally preventing iHP-140 from something like that being supported?
11:03:08GodEater_igorr: yes - the H1x0 hardware has no RTC chip
11:03:32igorrAh, so basically, no clock due to architectural limits of the player, then?
11:03:37GodEater_correct
11:03:49igorrThanks for the pointer
11:03:55GodEater_although there are some brave souls that have hardware mod'd their players to include this
11:04:01GodEater_and the process is documented in our wiki somewhere
11:04:19GodEater_in which case the RB build can also be modified to allow use of this component
11:04:55pixelmahttp://www.rockbox.org/twiki/bin/view/Main/RTCModH1x0
11:05:15igorrok, I admit −− the player is now a iHP-160, waiting to become a iHP-180 (with a new battery), but I am not sure I have the skills necessary for more invasive procedures :)
11:05:19GodEater_pixelma is our personal "search our wiki for you" service. Please tip her on the way out. :)
11:05:57pixelma:)
11:06:19igorrpixelma: whoa. Thanks for the tip (where is your tip jar, by the way?), but that modification is way out of my league.
11:06:22GodEater_igorr: allegedly it's not that tricky if you have some skills with a soldering iron, though I couldn't comment personally
11:06:31igorrI do not :(
11:06:39GodEater_know anyone who does ?
11:06:44igorrSure
11:07:03 Quit TotallyInfected (Remote closed the connection)
11:07:19pixelmamaybe you could try the Rockbox fund ;)
11:07:40igorrBut I think I will stick to replacing the hdd, getting a high-capacity battery and trying to get comfortable with the humongous number of features of the firmware :)
11:07:43 Join TotallyInfected [0] (n=ebola@pool-70-110-242-209.phil.east.verizon.net)
11:08:41igorrpixelma: bjorn@haxx.se?
11:09:33GodEater_igorr: that's Zagor - one of the original RB developers
11:09:35GodEater_he manages the fund
11:09:57igorrGoodie then. Do I need a paypal account, or can I just transfer funds from a credit card?
11:10:30GodEater_I think you need a paypal account too - but to be honest I don't remember =/
11:11:39pixelmaZagor is joined, he should be able to tell you more if he has the time
11:12:05Zagoryou can donate without an account. just enter the amount and then click "Use your credit card or bank account (where available). Continue" in the lower right of the page
11:13:04Zagornot very obvious, but then I guess they want you to sign up...
11:14:42***Saving seen data "./dancer.seen"
11:14:49igorrpixelma: ok then, you have been tipped :)
11:14:54pondlifeZagor: How's USB doing, btw?
11:15:03pixelmaigorr: thanks :)
11:15:14igorrGodEater_: I was a bit blind there. No paypal account necessary.
11:15:18igorrAnd thanks again for the pointers
11:15:24GodEater_no problem
11:16:02Zagorpondlife: slowly. it seems the chip protests everything I want to do with it :)
11:16:07pondlife:/
11:18:13pondlifeLinusN: Welcome back! I've been using the H300 wake-up alarm ( SVN bootloader + http://www.rockbox.org/tracker/task/7814 ) and it works well. Aside from the production of a new bootloader, do you know of anything holding up this being committable?
11:18:35pondlifehint :)
11:18:46LinusNpondlife: one small thing, peturs problem with his 80gb disk
11:18:58pondlifeAh, yes
11:19:09LinusNnot a problem if it is only him, but how do we know?
11:19:21pondlifeI guess nobody but petur can diagnose it either
11:20:09pondlifeMaybe make a V7 bootloader available, but leave V6 there in case? It's not as if it prevents use of the OF.
11:20:37LinusNmy thinking as well
11:21:02pondlifeUntil we do that, we probably won't get any more info on the issue.
11:24:13LinusNprobably no
11:24:36LinusNi discussed this with daniel last week, and we came to the same conclusion
11:28:20igorrbtw, another question −− does anyone have any tips for where one could get iHP-140 batteries?
11:28:34 Quit idnar (Nick collision from services.)
11:28:36 Join idnar_ [0] (i=mithrand@unaffiliated/idnar)
11:32:25 Join Frazz [0] (n=Fraser@thelawsons.plus.com)
11:33:48 Quit rvvs89 (Remote closed the connection)
11:35:07peturLinusN, pondlife: would it help if I attach my 80GB disk to my h320 and/or my old 40GB to my h380 to see if the problem is related to the disk or my player?
11:35:20LinusNpetur: good idea
11:35:23pondlifeGuess so
11:35:48peturI'll try tonight... real life is flooding my time schedule atm :/
11:35:50pondlifeDo the bootloaders share the same init sequence still?
11:36:52 Join afterwar [0] (n=IceChat7@bzq-219-183-200.static.bezeqint.net)
11:37:04LinusNpartly
11:37:28pondlifeNico_P: MoB is looking good, all known crashes seem resolved.
11:37:41LinusNthe h300 has a special sequence, because the hard disk power works a little differently
11:42:32 Join Thundercloud [0] (n=thunderc@resnet02.nat.lancs.ac.uk)
11:42:40 Quit amiconn (Nick collision from services.)
11:42:46 Join amiconn [0] (n=jens@rockbox/developer/amiconn)
11:44:54 Join barrywardell [0] (n=barrywar@host-194-46-226-143.dsl-ie.utvinternet.net)
11:58:46afterwarhow i open my g3
11:59:21 Quit TotallyInfected ()
12:00
12:06:03 Join agm3nt [0] (n=opera@bartek.tu.kielce.pl)
12:07:06GodEater_what's a g3 ?
12:07:14afterwariaudio g3
12:07:21JdGordonwith a g3 opener.... duh!
12:07:51GodEater_no idea - never seen one
12:08:44 Nick fxb__ is now known as fxb (n=felixbru@h1252615.stratoserver.net)
12:09:34GodEater_looks like you need nothing more exciting that a screwdriver though
12:09:41GodEater_from the pics I can find of it
12:10:06afterwarwhere i get screwdriver in that size
12:10:18GodEater_a tool shop ?
12:10:46afterwardo u have url ?
12:11:25afterwarI tried some small screwdrivers for it, but none fit
12:11:47GodEater_what sort of screw is it ? Does it have a special head? Or is it just a philips ?
12:11:59afterwarphilips
12:12:51GodEater_http://www.maplin.co.uk/ <−− where I got all my tiny screw drivers from
12:13:12afterwarthnka
12:16:21Nico_Ppondlife: nice :)
12:16:33Nico_Ppondlife: the no rebuffering bug is weird though
12:16:40pondlifeOnly happened once
12:16:51pondlifeI think it's related to enabling crossfeed during playback
12:16:56pondlifeI had some white noise too
12:17:00pondlifeBut rebooting fixed
12:17:09Nico_Pah I need to do some testing with crossfade
12:17:15GodEater_also running here with no problems
12:17:59pondlifeNico_P: Regarding the progress bar issue, do you retain the previous track's ID3 data until the WPS has moved on?
12:18:11Nico_Pthere's still a major bug in the USB handling. if you plug USB in, handles won't get deleted and playback can't resum
12:18:30pondlifePop a mention on the wiki..
12:19:51 Join Jeton [0] (n=chatzill@79.126.188.204)
12:20:02Nico_Ppondlife: I can't be sure of that
12:20:46pondlifeThat might explain some oddness I saw then
12:21:03Nico_Pwhat kind of oddness ?
12:21:19pondlifeWPS displaying rubbish for a second
12:21:34Nico_Poh ? when ?
12:21:35pondlifeOnly on H300, but I'm still researching
12:21:37pondlifeToday
12:21:54pondlifeLiterally for a second, at track transition
12:22:10pondlife(Or just before.)
12:23:26pondlifeI guess a call to audio_current_track() with a race condition?
12:23:27JetonNico_P: will MoB enable jpeg support for Album Art? And what do you mean in the wiki by "Clean implementation of Album art"?
12:24:06GodEater_Jeton: MoB has nothing to do with jpeg album art
12:24:07Nico_PJeton: jpeg is a different problem. by clean I mean "that doesn't spin the disk on every track change"
12:25:01GodEater_so hopefully will improve battery life somewhat ;)
12:25:03Jetoni see. So ne JPEG support in sight ? But still BMP will be supported in a more "clean" fashion .
12:25:14Jetonnice.
12:25:27Nico_PJeton: yes, that's it. jpeg probably won't happen anytime soon
12:26:00GodEater_not until someone writes is anyway :)
12:26:02Nico_Pit'd require having a jpeg decoder in the core, which is bloat
12:26:22Nico_PGodEater_: even if someone writes it, I doubt it'd get accepted
12:27:16Jetonwill this also enable better resizing for the Album Art? Now resizing makes the .bmp look jagged, not smooth.
12:27:21GodEater_true - but unless someone writes it, there's nothing to accept ;)
12:27:25LinusNi think support for embedded jpeg album art would be really nice
12:27:28Nico_Ppondlife: the problem with the track transition is the use of curtrack_id3
12:27:45GodEater_LinusN: embedded too ? You don't want much do you? :)
12:27:48Nico_PJeton: resizing is another problem too
12:27:55pondlifeI reckon there's such demand for JPEG Album Art, it might get into SVN. Maybe as a plugin call to do the decode and resize...?
12:28:30pondlifePersonally, I'd not use it, but I'd like to see the number of "custom" builds reduced ;)
12:28:31Nico_Ppondlife: a plugin would be fine, yeah
12:28:42LinusNi'd rather have it in the core
12:28:50pondlifeAnd AA seems to be one of the main drivers behind custom builds
12:28:58Jetoni think the demand for jpeg is that 'almost' all DAP's use JPEG for album art, and converting stuff from jpeg to .bmp .
12:28:59*LinusN goes to lunch
12:29:11Nico_PLinusN: really? I'm not against it but I know a few people who will be
12:29:14Jetonpondlife: you're right about that.
12:29:22LinusNNico_P: yes, really
12:29:43 Join hasmind [0] (n=harry@203.129.58.209)
12:29:56GodEater_how will embedded work anyway ?
12:30:05GodEater_does each track in an album have to include the artwork ?
12:30:06GodEater_or just one ?
12:30:19Nico_PGodEater_: it's embedded in each track
12:30:23GodEater_yuk
12:30:35pondlifeYuk indeed
12:30:44GodEater_what a waste of disc space
12:30:50pondlifeAnd buffer
12:30:52hasmindhey, i didn't think i was a noob with C until i saw the "rc->watever"
12:31:01hasmindwat is that?
12:31:05Nico_PJeton: most daps use a database for album art and convert the art to a more practical format
12:31:05hasmindwats the ->???
12:31:27Nico_Phasmind: rc->smthg is like (*rc).smthg
12:31:32GodEater_hasmind, dereferencing a pointer
12:31:35hasmindo right, ty
12:31:52JetonNico_P: yeah, but they use JPEG mostly (?)
12:31:59pondlifeI'd like cover.jpg in the folder if I was to use it.
12:32:03hasmindis that just normal C syntax?
12:32:06Nico_PJeton: they read the jpeg but don't use it internally
12:32:10GodEater_hasmind: yes very normal
12:32:14pondlifehasmind: Absolutely
12:32:20hasmindlolz kk
12:32:35hasmindguess i am a total noob
12:32:40Nico_PJeton: I'm pretty sure the conversion is done by the computer before transferring the track over to the DAP
12:32:49*GodEater_ would agree with Nico_P
12:33:02Jetonpondlife: that would be much better than embedded, or Track Name.jpg etc.
12:33:25*Jeton agrees with Nico_P too
12:33:51GodEater_pondlife: how would that work for database users though ?
12:34:05pondlifeEvery file lives in a directory, no?
12:34:25GodEater_would be interesting to see one that didn't :)
12:34:33hasmindgod is my name ACTUALLY irc? lame. damn Gaim
12:34:37Nico_Pwe could probably cache the resulting bitmaps to avoid having to do the decoding everytime
12:34:56pondlifeLet's get MoB done first... ;)
12:35:02Nico_Pindeed
12:35:03Jetonyou could also use TrackName.mp3 with TrackName.jpg for that matter.
12:35:19 Join Redbreva [0] (n=chatzill@host86-138-16-226.range86-138.btcentralplus.com)
12:35:21GodEater_Jeton: that's no longer album art then
12:35:22Nico_PJeton: the AA patch already has TrackName.bmp
12:35:24GodEater_that's track art
12:35:32Jetonoh.
12:35:47pondlifeTrack art is where we're headed though - album art is a special case of that
12:35:55Jetonwhy do they call it Album Art then? :P
12:36:07GodEater_pondlife: so you *do* want one pic per track in the buffer then
12:36:28Nico_Ppondlife: I'm thinking I should commit MoB soonish. once I'll have fixed the usb bug I'll add comments and start seriously considering committing it
12:36:40GodEater_yay
12:36:51pondlifeyayay
12:37:07markunJeton: because it's usually the cover of the CD/Album
12:37:15Nico_Pit mostly depends on you guys though... if you find showstopper bugs I'll refrain from committing ;)
12:37:27*GodEater_ hasn't found any bugs at all =/
12:37:51pondlifeNico_P: As long as you make sure you have a light week afterwards - may be a need for some ninja bugsquishing
12:37:55GodEater_pondlife is clearly better at breaking it than me
12:37:55 Quit J3TC- (Read error: 110 (Connection timed out))
12:37:58*Jeton is off. Later
12:38:02 Quit Jeton ("ChatZilla 0.9.78.1 [Firefox 2.0.0.8/2007100816]")
12:38:03hasmindtesting...
12:38:09hasmindyay im hasmind
12:38:39GodEater_I assume the usb bug won't show up on ipods Nico_P ?
12:38:44Nico_Ppondlife: hmm I'm going back home next week end for a week... I'll have a bit of time to code but less than ATM
12:38:51GodEater_since it effectively reboots ?
12:38:52Nico_PGodEater_: no
12:39:46Nico_Ppondlife: but I'm thinking it would be nice to commit before the week end so that people can start reviewing and maybe even helping
12:40:02Nico_PI'm hoping Slasheri will have things to say
12:40:08Nico_Pmaybe even lostlogic!
12:40:26SoapZagor: This _might_ be a minor security hole in the wiki:
12:40:47ZagorSoap: private, please
12:41:09pondlifeNico_P: I won't object
12:41:39markunNico_P: will you be arround a lot to help fix any serious bugs?
12:41:53Nico_Pmarkun: I sure won't disappear
12:41:58markungood :)
12:42:24markunit's always nice to commit something big just before going on a long holiday :)
12:42:26pondlifeNico_P: I think you should sort out the progress bar before commmit
12:42:48pondlifeAny ideas on how?
12:42:52Nico_Pyeah I keep forgetting about that :/ the fix isn't trivial
12:43:04Nico_PjhMikeS suggested timestamping the PCM buffer
12:43:28pondlifeThe info should come from pcmbuf.c, but a simpler fix might be possible
12:43:30amiconnDoes MoB fix the wps bug related to next track info, btw?
12:43:48Nico_Pamiconn: what bug exactly ?
12:44:18amiconnWell, current track metadata is synced to the track change at the pcm level, even if decoding already went ahead
12:44:50Nico_PI think that's the same issue we're discussing
12:45:05pondlifeNot quite...
12:45:18amiconnBut next track metadata isn't, so for around 2 seconds up to the end of a track, "next track" info contains info from 2 tracks ahead
12:45:26pondlifeAh, yes
12:45:33pondlifeSimilar issu
12:45:34pondlifee
12:46:16pondlifeplayback.c needs to move the current track along only when it gets the PCM callback to denote track change
12:46:30 Quit JdGordon ("Konversation terminated!")
12:46:49Nico_Pamiconn: the bug is most probably still there
12:47:25pondlifeNico_P: The solution to both bugs could be the same...
12:47:50Nico_Ppondlife: I agree, and I hadn't thought of the pcm callback... thanks for mentioning it :)
12:48:04 Join JdGordon [0] (n=Miranda@c210-49-113-143.smelb1.vic.optusnet.com.au)
12:48:10pondlifeThat's the first time that playback knows the track has changed.
12:50:47Nico_Ppondlife: it happens precisley when the track has changed at PCM level, right ?
12:51:46pondlifeI think so, yes
12:51:53pondlifeSim logfs?
12:51:55Nico_Pthat's what it seems
12:52:40Nico_PI think I could have a static prevtrack_id3 and make the WPS display that during the track transition
12:53:08Nico_Pand use the callback like in SVN to update it
12:53:21pondlifeYes
12:53:50pondlifeWhat about the next track problem?
12:54:21Nico_Phopefully it will be solved with the same fix
12:54:24pondlifeWhy not make the audio_current_track() and audio_next_track() return static copies at this point?
12:54:38amiconnIn theory you wouldn't need special measures. Just don't update track related wps info until you receive track change notification from pcm
12:54:41Nico_Pthey do, but not the right ones
12:55:00pondlifeamiconn: That's what I meant.
12:55:10pondlifeHopefully this could simplify code.
12:55:15Nico_Pamiconn: the problem is that the WPS uses curtrack_id3, which changes too eraly for the WPS
12:57:05pondlifeYou need to name the variables carefully - i.e. whether they are decoding or playback related.
12:59:13 Quit JdGordon (Read error: 104 (Connection reset by peer))
12:59:31Nico_Pthe codec and the WPS need to use the same struct mp3entry, otherwise the elapsed info doesn't get updated
13:00
13:00:46pondlifeGood point
13:01:05 Join JdGordon [0] (n=jonno@c210-49-113-143.smelb1.vic.optusnet.com.au)
13:01:17Nico_Pthe codec can't use the one in the main buffer because it can move around
13:01:54Nico_Pthat's why I was thinking of moving elapsed out of struct mp3entry, but it's a big change
13:02:16pondlifeDoesn't this apply to the whole structure really anyway - the timing I mean.
13:02:21Nico_Pand it's probably not even the best fix
13:02:26Nico_Pyes
13:03:00Nico_PI think the best and maybe only solution is to use a prevtrack_id3 struct and to manage it carefully
13:03:17pondlifeJust a placeholder for the transition window?
13:03:22Nico_Pyes
13:03:36pondlifei.e. a static to replace prev_ti
13:03:43Nico_Pexactly
13:04:10pondlifeOK, back later
13:08:21 Join rvvs89 [0] (n=rvvs89@martello.ucc.gu.uwa.edu.au)
13:14:38pondlifeNico_P: One thing I've noticed is that it seems codec transitions are much slower with MoB...
13:14:46***Saving seen data "./dancer.seen"
13:14:54pondlifeCrossfade from MP3 to WMA is not smooth
13:14:58pondlife({or back)
13:15:31hasmindis there a crossfade setting?
13:15:41pondlifeThere are many!
13:15:44hasmindhaha
13:15:46Nico_Ppondlife: OK I'll try to have a look... how is it with the same codec ?
13:15:46hasmindwhere?
13:16:02pondlifeNico_P: Fine, very quick track transitions in general
13:16:09pondlifePossibly better than SVN
13:17:00Nico_Peven with crossfade on ?
13:17:31pondlifeYes
13:17:39Nico_Pcool
13:17:54Nico_PI'll have a look at the codec transitions
13:18:06Nico_Phmm even in SVN the playlist index is updated too early
13:18:57pondlifeWooh, White Noise
13:19:22pondlifeOK, set up a slow-ish crossfade, but leave crossfading disabled
13:19:38pondlifeThen enable it during playback (which stops playback, a bug that's already in SVN).
13:19:49pondlifePress play, and wait a while - white noise!
13:20:06GodEater_define slowish
13:20:07Nico_P:/
13:20:13pondlife5 seconds
13:20:29GodEater_at both ends ?
13:21:01pondlifeAh no. Try params: 0 / 0 / 1 / 1 / Crossfade
13:21:04pondlifeStill does it
13:21:44GodEater_0011 ?
13:21:49 Part hasmind
13:22:01GodEater_no fade in delay, no fade in duration ?
13:22:09GodEater_and then only 1 second of each to fade out ?
13:22:11pondlifeCorrect
13:22:26pondlifeStraight in with the new track, quick fade the old
13:22:39pondlifePerfectly valid...
13:22:39 Quit B4gder ("It is time to say moo")
13:22:53pondlifeMight only give white noise after a codec transition too.
13:23:12pondlifeI have a test playlist alternating MP3 and WMA
13:23:22GodEater_I'm getting no noise at all
13:23:40 Quit barrywardell ()
13:24:05safetydanpreglow, the uwb speex file I have locks up when played with the latest svn
13:24:15GodEater_and then trying to rebuffer, I've just got "loading"
13:24:19*safetydan goes digging for a paper clip
13:24:20GodEater_and it won't go away
13:24:27pondlifeThat's what happened to me earleir
13:24:35pondlifeWhen I could spell :)
13:24:55GodEater_yep - complete hard lockl
13:25:09pondlifeI could press STOP, but had to reboot to get playback
13:25:13preglowsafetydan: and it didn't use to lock up?
13:25:24*preglow really wouldn't mind seeing jpeg support in the core
13:25:37safetydanpreglow, no these were test files that all played successfully before, but I'll verify that
13:26:02preglowit would add what, 10kb to the binary? i think it would be worth it on the targets where user would expect it to work
13:26:04 Quit afterwar ("I cna ytpe 300 wrods pre mniuet!!!")
13:27:24Nico_Ppondlife: I've manage dto get something good going
13:27:53preglowsafetydan: i've got a hunch what's wrong
13:27:56GodEater_hmm - doing anything with crossfade appears to break playback
13:28:19Nico_PI haven't done any real testing with crossfade :/
13:29:28safetydanpreglow, yeah all good with a build before your changes
13:29:29pondlifeI recommend you do!
13:29:39preglowsafetydan: yeah, i've found the bug
13:30:07pondlifeTo be honest most of the crossfade problems I'm seeing are a result of changing settings during playback rather than crossfade per se.
13:30:55pondlifeWooh! Resumed from reboot and now I have wacky times on the WPS. Total track time −−293:-45 (and counting down)
13:31:13pondlifeCurrent position: 4:57:51 (in a 2 minute track)
13:31:53 Join barrywardell [0] (n=barrywar@host-194-46-226-143.dsl-ie.utvinternet.net)
13:32:55pondlifeNico_P: After a reboot, the buffering debugging screen shows track_count: -1. Is this correct?
13:33:06pondlifeHandle count = 0, as expected.
13:33:08Nico_Ppondlife: yes
13:33:13pondlifeOK
13:34:57preglowsafetydan: like i thought, uwb mode really uses TWO wb decoder state structs, so my current static allocated one will be used twice
13:35:00preglownow, how to deal with this
13:35:48amiconnpreglow: The problem with jpeg decoding is that you need a buffer to decode into. And unlike with preconverted BMP album art, you don't know how big an image the track will have embedded
13:35:57safetydanpreglow, cheat and change speex_alloc to work out of a buffer in iram?
13:36:32preglowsafetydan: i guess that would work
13:36:35amiconn..and embedded jpegs should *really* not be called album art, imo
13:36:52preglowamiconn: i agree there, but it is a solution that exists :/
13:37:12preglowand i am as usual opposed to people having to alter their files for rockbox to be able to use them
13:37:13*amiconn thinks of it more as a problem than a solution :(
13:37:20preglowanyway
13:37:27preglowall the album art i have seen has been in jpeg
13:37:30pregloweven if it isn't embedded
13:37:40amiconnAnd imo jpeg decoding should not be part of the core
13:38:01amiconnWhy wasting binsize if a user never wants to use album art?
13:38:13preglowamiconn: why is the image size a problem, btw? we'll have mob
13:38:17pixelmaI've seen other formats there as well (png IIRC) in my very few tracks that has them
13:38:21amiconnIt could be a plugin that decodes the jpeg and shows it in a viewport on the wps
13:38:25preglowpixelma: blerhg :/
13:38:49Nico_Pamiconn: I just fixed the next track info bug in SVN... the playlist index is still incremented too early though
13:38:55Nico_Perr not SVN, MoB
13:38:58amiconnpreglow: Do you really want to store the *decoded* jpeg in the playback buffer??
13:39:23Nico_Pamiconn: decoded an resized, why not ?
13:39:38amiconnResized? __where__???
13:40:02amiconnYou can predecode at 1/2, 1/4, or 1/8, but that's it
13:40:20preglowmaking a simple resizing function is a piece of cake
13:40:28LinusNamiconn: in the same way we resize the bmp's today i guess
13:40:37Nico_PLinusN: we do?
13:40:48LinusNin the bmp resize patch
13:40:55Nico_Phahaha
13:40:56amiconnLinusN: We resize BMPs? Not that I know of
13:41:36preglowanyway, that's not a problem
13:41:39amiconnBMP resizing is way easier than resizing jpeg. BMP can be resized on load because it's uncompressed
13:41:40preglowit won't even each much bin size
13:41:53pregloweat...
13:41:57amiconnAnd btw, the bmp resizing patch is a crappy implementation
13:42:00 Quit safetydan ("Leaving")
13:42:02LinusNsure it is
13:42:13amiconn(or at least was the last time I looked at it)
13:42:37amiconnpreglow: You need some buffer to work on...
13:42:57preglowamiconn: yeah, but not a full size one
13:43:11preglowamiconn: album art files usually aren't 1600x1200
13:43:24preglowyou can resize as you decode, even if you don't do it the way we currently do
13:44:42pondlifeFor non-embededed art it would be reasonable to put a small limit on the dimensions anyway
13:44:43preglowthe core will always waste bin size on features some people will never use
13:44:48preglowand quite some people WILL use jpeg album art
13:44:54pixelmapreglow: 3 pngs, 11 jpgs - I extracted them and deleted them all in the mp3
13:45:25preglowsilly people...
13:47:21*preglow tries to figure out if he likes the speex_alloc() out of iram idea...
13:47:45 Quit Rob2222 (Read error: 104 (Connection reset by peer))
13:49:05 Join Arathis [0] (n=doerk@p508A650D.dip.t-dialin.net)
13:50:24Nico_Ppondlife: committed
13:50:33Nico_P... and pushed too
13:54:29 Quit Xerion (Read error: 104 (Connection reset by peer))
13:55:24 Join Xerion [0] (i=xerion@cp198589-d.landg1.lb.home.nl)
13:56:08 Quit barrywardell ()
13:56:08*Nico_P is off to class again
13:56:28 Quit Nico_P (Remote closed the connection)
14:00
14:02:02 Join Mouser_X [0] (n=someone@207.155.176.3)
14:02:35*preglow shaves 40kb off speex.codec
14:06:42preglowdoes sectioned functions have to be enabled for the linker to remove unneeded stuff?
14:06:58preglowwouldn't really work in this case, but...
14:07:28 Join lee-qid [0] (n=liqid@p54966122.dip.t-dialin.net)
14:14:36 Join Rob2222 [0] (n=Miranda@p54B1671C.dip.t-dialin.net)
14:18:18 Join Genre9mp3 [0] (n=yngwiejo@athedsl-84948.home.otenet.gr)
14:20:14*JdGordon is touchpad king!
14:22:09*preglow bows to the king of touchpad
14:23:46 Part LinusN
14:24:54 Quit kubiix (Read error: 104 (Connection reset by peer))
14:31:05 Join spiorf [0] (n=spiorf@host98-168-dynamic.6-79-r.retail.telecomitalia.it)
14:33:38 Quit XavierGr ()
14:34:09 Join egolost [0] (i=egolost@soll3-41.cust.blixtvik.net)
14:34:34egolostwhat is the smallest mp3 player that support rockbox or another linux firmware?
14:34:55JdGordoneither the sansa c200 or the nano
14:35:40egolostbut sansa is better supported?
14:35:55JdGordonthey are supported the same
14:36:22JdGordonthe sansa has the sd slot which the nano doesnt ...
14:36:46pixelmathe _c200_ it's quite a new port - the only big thing missing is the microSD support
14:36:59pixelmahopefully to come soonish
14:37:00JdGordonmsd isnt working on the c200 yet?
14:37:07pixelmanope
14:37:18JdGordonoh.. /me thought it was :p
14:37:21egolostis it c2xx or just the c200?
14:37:42egolostcause i'm thinking about a c250R.
14:37:45 Join PaulJam [0] (i=PaulJam_@vpn-3104.gwdg.de)
14:37:51JdGordonDONT buy an r
14:38:02pixelmawith c200 we mean the series, there is no c200
14:38:36pixelmaI didn't even no that there was also an R version... *puzzled*
14:38:43pixelma*know
14:39:04egolostSo, the radio isn't supported then? :D
14:39:24pixelmathe radio is supported
14:40:02egolostso why should'nt I buy an R?
14:40:32JdGordonbecause if its anything like the e200R its a pita to get rockbox on it
14:40:41JdGordonand by pita, i mean we dont know how to do it
14:41:07pixelmabecause you caused us confusion... with the Sansa e-series there are the so-called "Rhapsody" models which are called e250R for example
14:41:20JdGordonthe e200r was a fair few months behind the regular e200 to get rockbox
14:41:44pixelmaI have a c250 with radio and the box only says "c250" nothing more
14:42:22 Join nicktastic [0] (n=nick@unaffiliated/nicktastic)
14:42:47pixelmaso just a misunderstanding with the name...
14:42:54 Join XavierGr [0] (n=xavier@ppp176-207.adsl.forthnet.gr)
14:43:54egolostis there sansa c250R with rapsody?
14:44:00*JdGordon wonders what the heck the c250r is... google only brings hits from .sk and .cz
14:45:04amiconn[14:34:34] <egolost> what is the smallest mp3 player that support rockbox or another linux firmware? <== fyi: Rockbox is *not* linux based
14:45:05 Join B4gder [0] (n=daniel@rockbox/developer/bagder)
14:45:33B4gderthere are random talks about a Rhapsody c200 model, called "v2" at places
14:48:05 Quit Sedgewick (Read error: 110 (Connection timed out))
14:59:16pixelmaI found a german webshop that has a "c250R" and a "c250 with radio" (the latter is like 4 eur. cheaper) but I can't figure out what the difference should be. Besides that there info is a bit wrong anyways, they say their display is 128x96...
14:59:33pixelmas/there/their
15:00
15:01:03preglowwhat size is the c200 lcd, again?
15:01:21pixelmahaha, the description text says it would have an fm tuner but the table doesn't...
15:01:28pixelmapreglow: 132x80
15:01:45preglowthat sure isn't much
15:01:48preglowbut it looks cute :)
15:02:08krazykitoh, my e200 battery bench is done! i should check it and add the info today
15:02:20preglowi wonder how mpegplayer looks on that thing
15:02:34pixelmait's not that bad
15:02:45egolostthis is the one I'm considering buying.. if it runs rockbox that is : http://www.webhallen.com/prod.php?id=67646
15:02:56JdGordonkrazykit: about 15h?
15:03:11krazykitJdGordon, i don't know. refresh database :-(. i need to build the svn bootloader
15:05:52krazykitwhat the crap. 4 hours? that can't be right. i KNOW it gets more than that.
15:07:01krazykithm, the battery claims to start at 50% full, which is strange
15:08:14 Join Sedgewick [0] (n=10334AEF@host99-233-dynamic.15-87-r.retail.telecomitalia.it)
15:08:55krazykitit's strange: the device claimed "low battery: shutting down" when i turned it on, and the 2 minutes of being plugged in, rockbox now reports 49% charge
15:09:16pixelmaegolost: if the mean radio with the "R" then yes this one should run rockbox
15:09:41pondlifeJdGordon: Did you enable touchpad/mouse support for all sims in the end?
15:11:20krazykitin 3 minutes, the battery has gone from 49 to 38%. i think this battery is shot.
15:11:36JdGordonpondlife: not yet
15:12:16pondlifeOK, I'm not convinced that you should do!
15:12:46*JdGordon doesnt see the harm in it.. but probably wont enable it anyway
15:13:31pondlifeIn case we end up using #ifdef HAVE_TOUCHPAD in more places to decide which UI to use... the sim should match the target as closely as possible
15:13:43JdGordonyeah, true
15:14:49***Saving seen data "./dancer.seen"
15:27:42 Join webguest78 [0] (i=a8d89499@gateway/web/cgi-irc/labb.contactor.se/x-1f9fb898dfcf12f2)
15:30:20 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
15:30:25 Nick fxb is now known as fxb__ (n=felixbru@h1252615.stratoserver.net)
15:30:30 Quit Rob2222 (Read error: 104 (Connection reset by peer))
15:31:09 Part webguest78
15:32:02 Quit Frazz (Read error: 104 (Connection reset by peer))
15:40:30 Quit Gnu47 (Remote closed the connection)
15:40:42 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
15:44:03 Join kugel [0] (i=kugel@unaffiliated/kugel)
15:48:07 Join n1s [0] (n=nils@nl104-209-90.student.uu.se)
15:49:46 Join desowin [0] (n=desowin@hdp186.internetdsl.tpnet.pl)
15:50:52 Quit Gnu47 (Remote closed the connection)
15:51:05 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
15:55:08 Join jgarvey [0] (n=jgarvey@cpe-069-134-102-044.nc.res.rr.com)
15:57:28 Part Redbreva
15:58:31*pixelma wonders why commited r.15267
15:58:36pixelma*who
15:59:32n1sJdGordon: did
15:59:54pixelmayeah, found it
15:59:58*n1s pats svn log
16:00
16:00:05JdGordonwhats 15267?
16:00:29pixelmaI just wondered how one managed to put that in between my 2 commits...
16:00:32 Quit Mouser_X (Read error: 110 (Connection timed out))
16:00:42JdGordonah
16:01:15 Quit Gnu47 (Remote closed the connection)
16:01:16n1spixelma: how did the midi test go last night?
16:01:27 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
16:02:48n1ssorry for disappearing...
16:03:13pixelman1s: didn't get data aborts on the c200 with the one file if that's what you mean :)
16:03:32n1swas that the file that gave data aborts before?
16:03:51pixelmathe one I sent you still doesn't play right
16:04:14n1spixelma: yeah, that is some other issue...
16:04:32 Join J3TC- [0] (n=jetc123@dhcp75-19.njit.edu)
16:05:20pixelmaI tested with the midi that caused data aborts but this is not the one I sent you
16:05:50n1sso the data aborts are fixed?
16:05:59*n1s is a bit slow today
16:06:19pixelmawouldn't have made much sense because it was well on my M5 even before... yes, seems to be fixed now :)
16:06:24 Join Nico_P [0] (n=nicolas@rockbox/developer/NicoP)
16:06:50J3TC-:O
16:06:54J3TC-I'm at history class
16:07:02pixelmaoops
16:07:04J3TC-Compiling rockbox
16:07:06J3TC-:3
16:07:20n1sgood, there was a bug that caused memory corruption with certain files because I forgot to check the size before writing to memory...
16:08:44 Quit B4gder ("It is time to say moo")
16:08:53pondlifeNico_P: Any idea how we can detect the user rewinding during the last seconds of a track?
16:09:09Nico_Pnot really
16:09:20Nico_Pmaybe with the wps_offset
16:09:37Nico_Pthe WPS not displaying the right track after the seek is a refreshing problem
16:10:13pondlifeI was thinking you might be able to do something in the same window now you have a test anyway
16:10:37Nico_Psame window ?
16:11:29 Quit qwm (Remote closed the connection)
16:11:36 Quit Gnu47 (Remote closed the connection)
16:11:40 Join qwm [0] (n=qwm@h38n2fls32o1010.telia.com)
16:11:49 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
16:11:56Nico_PJdGordon: regarding mouse in the sim I still think a command line switch would be nice
16:12:27pondlifeI mean the window between decode and WPS track transitions.
16:12:46pondlifeCurrently, a rewind in that window results in the WPS being one track out
16:12:53Nico_Pah, right. well yeah, now it's quite well defined
16:12:54pondlifeUntil you stop and restart playback
16:13:16 Quit GodEater_ (Read error: 110 (Connection timed out))
16:13:16pondlifeSo I'm not sure it's just a refresh issue
16:13:22Nico_Ppondlife: it's just a refresh problem. show another screen then return to the WPS and it's fixed
16:13:31 Nick fxb__ is now known as fxb (n=felixbru@h1252615.stratoserver.net)
16:13:33Nico_Punless that changed with my latest commit
16:14:14pondlifeWell if I get in that state, then go to the settings screen and back it remains wrong
16:14:24pondlifeIt stays wrong over track transitions too
16:14:34pondlifeOnly stopping completely works
16:14:49pondlifeI thought that was how it had always been.
16:15:02Nico_Pah yes, that's new to my commit and it's probably due to the use of wps_offset
16:15:17Nico_PIIRC before it was less serious
16:15:49pondlifeHmm, at least it's consistent now
16:16:15Nico_Pyeah, not sure it's better :s
16:16:31Nico_Pwell I know what code isn't executed
16:17:23Nico_PI'm starting to feel the need to rethink playback.c
16:17:57Nico_Pbut I have no idea how I would do it so it's not very productive...
16:18:07pondlifeI'd wait until MoB is in
16:18:21pondlifeAlthough whilst you have the knowledge that would be nice
16:18:23Nico_Pof course
16:18:56pondlifeEven your latest one does the track transition too early in the playlist viewer
16:19:11Nico_Pyeah what I want to do is commit when there are no major regressions and then maybe refactor more deeply... what I've done up till now is basically a translation
16:19:15pondlifei.e. it moves down the list at the decode transition
16:19:36Nico_Pyes, that is consistent with the playlit index being updated too early
16:19:47Nico_Pmaybe I can improve that though
16:20:18pondlifeWe need to make all external transition happen when triggered by the PCM callback
16:20:25 Quit JdGordon (Read error: 110 (Connection timed out))
16:20:26Nico_Pthe problem is that most of the track transition is handled by audio_check_new_track, which is triggered by the codec requesting a new track
16:20:39pondlifeAh, that's the DECODE transition
16:20:52pondlifeNeed to go though that code and split it up?
16:21:17pondlifedecode transition = internal only
16:21:59 Quit Gnu47 (Remote closed the connection)
16:22:09 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
16:22:09Nico_PI'll try to move the playlist_next() call, even though I don't really know how important it is
16:22:44pondlifeThe main problem is that some callbacks occur in an interrupt...
16:22:57Nico_Pwhich ?
16:23:03pondlifeNot sure
16:25:18 Quit Zagor ("Client exiting")
16:27:21 Part agm3nt
16:27:32Nico_Ppondlife: have you by any chance noticed gapless breaking ?
16:27:38 Join mrkiko [0] (n=mrkiko@adsl-ull-185-119.42-151.net24.it)
16:27:45mrkikoHi all!
16:28:02pondlifeNico_P: Not tested it
16:28:04Nico_Pok
16:28:18Nico_Pit's probably just the tracks I listened to
16:28:26Nico_PI'll double check of course
16:28:28mrkikoSince I should do many homeworks, if someone has something to tell me and would like to be sure I'll read it, he is pleased to open a private window. thank!
16:29:58pondlifeNico_P: Breaking news, I just had playback stall again - i.e. it failed to rebuffer.
16:30:17Nico_P:/
16:30:25pondlifeI'm in the debugging screen - useful = 8059 bytes and not decreasing.
16:30:26Nico_Pwhat kind of situation was it ?
16:30:36pondlifeRight in the middle of a track
16:30:49Nico_Pand how full is the buffer ?
16:30:56pondlifealloc = full
16:31:15Nico_Pthat's very strange
16:31:18pondlifereal = 345k
16:31:26pondlifeusefl = 8059 bytes
16:31:44pondlifeIt's just sitting there, not locked but not buffering
16:31:57Nico_Pyeah so the codec is starving
16:31:59pondlifeAnything you'd like me to try?
16:32:10pondlifeI suspect your !buffering test...
16:32:15mrkiko... has rockbox allways been so crashful?
16:32:18Nico_Pmaybe skip backwards. you can do it by pressing up in the debug screen
16:32:20 Quit Gnu47 (Remote closed the connection)
16:32:20pondlifei.e. race condition
16:32:35pondlifeNico_P: Yes, that rebuffered
16:32:48Nico_Ppondlife: did "buff:" have an y next to it ?
16:33:03pondlifeSorry, I didn't see
16:33:08Nico_Ppondlife: it shouldn't have rebuffered, just gone back to the prev track which should be in mem
16:33:36Nico_Pie normally usefl should have increased without the disk starting or anything
16:34:07pondlifeAh, well I saw an increase in real too, so that indicates buffering, yes?
16:34:24pondlifeNow real is back up to 25MB
16:34:31Nico_Pah yes if real or alloc change, that means buffering is happening
16:34:51Nico_Ppondlife: I'm thinking maybe the !pcmbuf_is_lowdata() condition is the cause.
16:35:05pondlifeHmm, PCM was at 0
16:35:30Nico_Pyeah so refilling wasn't about to happen. strange the pcm buffer was already low when data was first needed though
16:36:09Nico_Pwhat codec was it ?
16:36:20pondlifeAh, it won't rebuffer if PCM is starved so as to allow it to refill? That should only be true if there's useful data...
16:36:23pondlifeMP3
16:36:24kugeldo I need to build the bootloader without checking out the whole source
16:36:55kugeli checked out /trunk/bootloader, but make doesn't work
16:37:40Nico_Pkugel: I think the bootloader relies on firmware/
16:37:55 Join JdGordon [0] (n=jonno@c210-49-113-143.smelb1.vic.optusnet.com.au)
16:37:56kugelah, damn
16:38:08Nico_Pyou can just check that out though
16:38:42kugelthat's pain with cygwin
16:38:44kugelimo
16:38:48Nico_Ppondlife: I now have the playlist index being incremented when pcm changes tracks
16:39:27kugeldoes anyone know what "No original firmware found at 0xeed15200" means?
16:39:39kugelthat's the error when I do sansapatcher -of OF.mi4
16:39:40*pixelma hopes that everything's back to normal now... :\
16:41:23pondlifeNico_P: Would you agree with me that to the outside world, the "WPS track transition" should be the only one?
16:41:24Nico_Ppondlife: I think I'll simply disable seeking in the transition window
16:41:35Nico_Ptotally
16:41:59pondlifeNico_P: You might find that has more side-effects than fixing the problem ...
16:42:18 Quit pill (Nick collision from services.)
16:42:35 Join pill [0] (i=pill@sloth.shellfx.net)
16:42:43 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
16:42:49Nico_Ppondlife: why? to me it makes sense. during the transition window we're already in the new track. seeking in the WPS-current but codec-previous is impossible
16:43:16Nico_Punless we skip back again and then seek
16:43:38 Quit pill (Nick collision from services.)
16:43:46pondlifeYes, and with your new algorithm, the chance is the track will be buffered still
16:43:50 Join _pill [0] (i=pill@sloth.shellfx.net)
16:43:52pondlifeSo it should work smoothly
16:44:03pondlifeAh, just stalled again
16:45:56pondlifepcm = 0/882000 handles=4 tracks=1, data_rem=1907759, alloc=25,164,228/25,191,824, real=2,517,976/.., usefl=8166/...
16:46:01pondlifebuff:
16:46:04pondlifei.e. no y
16:46:57Nico_Pare you managing to make it happen ?
16:47:12pondlifeI just left it playing
16:47:35pondlifeI skipped one track a while back, but otherwise just let it play
16:47:38Nico_PI think the !pcmbuf_is_lowdata can and should go
16:47:50pondlifeNot so sure.
16:48:03pondlifeIt should have triggered the rebuffer earlier anyway.
16:48:18Nico_PI agree but I really see no reason why it didn't
16:48:28pondlifePCM lowdata would only have occurred a second or so beforehand anyway I expect
16:49:08pondlifeCould you perhaps remove the !buffering? I can try it here if you want...
16:49:18Nico_Pyeah why not
16:49:32Nico_Pwhere in the playlist is the track? not last ?
16:49:53pondlifeNope
16:49:56pondlifeNothing special
16:50:09pondlifeAgain, we were mid track
16:50:16pondlifeNowhere near a transition
16:50:39Nico_Pyeah but it's probably simply the first track that isn't completely buffered
16:50:45 Join scorche|w [0] (n=8dc5049d@ircby.iptel.by)
16:50:53pondlifeIndeed, it is
16:51:28pondlifeHmm, is audio_initialize_buffer_fill() needed still?
16:51:36pondlifeAnd the filling variable?
16:51:47Nico_Pnot quite sure actually
16:51:52pondlifeI would think that complexity should have vanished into buffering
16:52:40pondlifeaudio_fill_file_buffer() is still called lots
16:52:54 Quit Gnu47 (Remote closed the connection)
16:53:08 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
16:53:21Nico_Pyeah that's why I was saying I want to refactor the playback code... this one should basically disappeatr
16:53:58pondlifeAbsolutely. playback.c should end up with only the playback in (i.e. reading from the buffer)
16:54:42 Join GodEater_ [0] (n=bryan@rockbox/staff/GodEater)
16:55:14Calcipherpondlife, hows it going
16:55:24Nico_PI'm really pleased I used git for all this. I would've gone crazy without it
16:55:39pondlifeCalcipher: Not that well :)
16:56:42 Join barrywardell [0] (n=barrywar@dhcp-892b9af6.ucd.ie)
16:56:52Nico_Poh? what's wrong?
16:57:11pondlifeBusy Real Life mainly
16:57:49pondlifeOK, buffering check removed - testing continues...
16:58:17Nico_Pit doesn't seem to happen on my sim. trying on the gigabeat
16:58:41Nico_Phave you tried using the debug screen to skip to the first partial song and waiting to see if it rebuffers ?
16:58:47pondlifeWill do
16:59:01Nico_Pit's most probably in that case that the problem happends
17:00
17:00:47pondlifeOK, it rebuffered ok that time
17:01:11Nico_Pfor me too... did it happen with the same track each time ?
17:01:18pondlifeNo
17:01:31mrkikois steve here?
17:02:14 Part Calcipher
17:02:47pondlifeOK, and worked ok again.. I think that !buffering test should go
17:02:53Nico_Pit's a pity the scrollbar widget can't display a ringbuffer. it's be nice to have in the debug screen
17:02:56JdGordonoy joy! looks like i got battery measurements working, but broke the touchscreen values :(
17:03:17Nico_Ppondlife: I have the !buffering test and it's not hurting
17:03:18 Quit Gnu47 (Remote closed the connection)
17:03:26pondlifeNico_P: How does A-B repeat display work?
17:03:29Nico_Ppondlife: I don't see why it would hurt
17:03:31 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
17:03:40pondlifeRace condition..
17:03:47Nico_Phow?
17:04:04Nico_Ppondlife: it displays some triangles. but I would like to have exactly like a scrollbar except it could wrap
17:04:04pondlifeNot sure, but you shouldn't access the buffering thread variables
17:04:35pondlifeThread safety is not to be trifled with, I'm afraid.
17:04:58Nico_Pont thread writes the var, others read it... isn't that safe ?
17:05:13Nico_Ps/ont/one
17:07:26pondlifeMay be, but there's always a chance for threads to lock each other out if there are further dependencies.
17:07:38pondlifeBest to keep them totally independent.
17:07:50Nico_Pyeah
17:08:08pondlifecodec_advance_buffer_counters() has a similar test
17:08:30pondlifeProblem is, that is in the codec thread, deciding if it's worth trying buffering.
17:08:39Nico_Pthat's where I thought you had removed the !buffering test, where have you?
17:08:45pondlifeWe should let the buffering thread have a chance to decide
17:08:49 Quit PaulJam (".")
17:08:51pondlifeThe other one, that you put in recently
17:09:13Nico_Pok
17:09:14pondlifeIn audio_thread
17:09:31pondlifeI'd make buffering a static bool
17:09:35pondlifeHide it away
17:09:46pondlife(Maybe include it in the debug info.)
17:10:07Nico_Pit's of no use if it's static. I put it there only so that other threads know whether buffering is in progress
17:10:20pondlifeOK, scrap it!
17:10:21Nico_Pyeah it's only be useful for debugging
17:10:36Nico_Pbut I need it! :p
17:10:50pondlifeOK, you need it, but playback.c doesn't :p
17:11:38Nico_Pthe best would of course be to remove the Q_AUDIO_FILL_BUFFER business and maybe give the buffering thread a callback to signal data is low
17:12:04pondlifeAbsolutely
17:12:04Nico_Phmm I got confused but I mean "do things differently)
17:12:47pondlifeActually, shouldn't the buffering thread deal with this automatically, assuming playback yields..
17:12:51Nico_PI've just pished my commit for the playlist position
17:13:04pondlifei.e. why need to have any artificial control
17:13:10pondlifeThanks
17:13:13Nico_Pyou mean the buffering thread could deal with the playlist directly ?
17:13:33pondlifeNo
17:13:42 Quit Gnu47 (Remote closed the connection)
17:13:46pondlifeI mean it should realise that the buffer is low automatically
17:13:55 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
17:14:07pondlifeAh, it doesn't know which file to get, does it?
17:14:21Nico_Pthe problem is that there are two different definitions of low buffer. low in files or low in buffered data
17:14:36pondlifeLow in files shouldn't be relevant much
17:14:42pondlifeIt's low in data that matters
17:14:46Nico_PI agree
17:14:54***Saving seen data "./dancer.seen"
17:15:05Nico_Pbut currently the audio thread actively tries to fill the buffer with files.
17:15:21Nico_Pmaybe I should change it so that it provides files when requested by the buffering thread
17:15:31pondlifeYes, that would be better
17:15:52Nico_PI'll try implementing that
17:17:57Nico_Ppondlife: I'm thinking users of the API could register callbacks, kinda like with register_ata_idle_func
17:18:14Nico_Pand those get called by the buffering thread when the buffer is low
17:20:35pondlifeAre you going to scrap those !buffering tests?
17:20:43*pondlife fears git non-synced-ness
17:21:11Nico_Pyou want me to scrap them now before implementing a proper solution?
17:21:25Nico_PI wasn't really planning on it
17:21:26pondlifeWell, I'm fairly sure they're not required
17:21:36pondlifeI don't see what good they do
17:22:02pondlifeThey may save a bit of calling, but nothing that would fail.
17:22:36pondlifeAnd I've had 3 successful rebuffers since removing them
17:22:54Nico_PI'll test without them
17:23:17pondlifeOnly the buffering thread should decide if there's buffering to do.
17:23:52Nico_Pin the audio thread it's used to decide if it's useful to start file filling
17:24:03kugeldid anyone read my question i asked about an hour ago?
17:24:05 Quit Gnu47 (Remote closed the connection)
17:24:06kugeldoes anyone know what "No original firmware found at 0xeed15200" means? that's the error when I do sansapatcher -of OF.mi4
17:24:14Nico_Pit's useless to do file filling if the buffering thread is filling
17:24:15pondlifeYou have "filling" for that
17:24:18 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
17:24:19 Join japc [0] (n=japc@194.65.73.69)
17:24:27pondlifei.e filling is for file filling, buffering for buffering.
17:24:37pondlifeNot the same thing, timing wise
17:24:42Nico_Pfilling is different
17:24:59Nico_PI do agree a better solution is needed anyway
17:26:13 Quit J3TC- (Read error: 110 (Connection timed out))
17:27:19Nico_Ppondlife: done
17:27:41Nico_Pwhat do you think of the callback idea?
17:29:31kugelnevermind guys, I solved the problem
17:31:26preglowhrmf
17:31:35preglowdoesn't look like speex asm will help us much
17:34:29 Quit Gnu47 (Remote closed the connection)
17:34:39 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
17:36:54 Join Rondom [0] (n=Rondom@p57A95747.dip.t-dialin.net)
17:37:41 Nick idnar_ is now known as idnar (i=mithrand@unaffiliated/idnar)
17:38:38 Quit JdGordon ("Konversation terminated!")
17:44:51 Quit Gnu47 (Remote closed the connection)
17:45:04 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
17:48:49 Quit scorche|w ("CGI:IRC (EOF)")
17:49:08 Join Calcipher [0] (n=Calciphe@ool-18bab657.dyn.optonline.net)
17:51:43pondlifeNico_P: I like the callback idea
17:52:08pondlifeOne minor bug (which might be in SVN too)... You shouldn't be checking !filling before putting Q_AUDIO_FILL_BUFFER in the queue
17:52:15egolosthmm.. seems like my sandisk sansa c250 hangs when i plug in into the USB.
17:52:20pondlifeThat test is done on the audio thread anyway
17:52:21egolostwith rockbox.
17:52:36Nico_Ppondlife: hopefully I'll be able to get rid of it
17:52:55pondlifefilling is now an audio-thread only var anyway
17:53:55pondlifeSo things are definitely clearing up lots
17:53:58pixelmaegolost: that happens with some builds and you plug in usb with rockbox running. Try switching it off and then connect usb, your sansa should boot into the original firmware
17:55:02 Join J3TC- [0] (n=jetc123@dhcp75-19.njit.edu)
17:55:14 Quit Gnu47 (Remote closed the connection)
17:55:19 Join Davide-NYC [0] (n=chatzill@user-12hdtj8.cable.mindspring.com)
17:55:26 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
17:55:32egolostpixelma: thank you.. that seems to work fine.
17:55:37preglow215% -> 220% for an ultra-wideband 32khz file with iir_mem16 asmified, very disappointing :/
17:56:18pondlifeNico_P: I suspect today was the day MoB got better than S
17:56:20pondlifeVN :)
17:56:25Nico_P:)
17:56:49pondlifeOn the pondlife-entirely-subjective scale, of course
17:57:35 Join scorche|w [0] (n=8dc5049d@st.iptel.by)
17:57:45J3TC-Which file is the iaudio rockbox boot image?
17:58:22Davide-NYCjhMikeS: could the recent thread_wait commit solve the test_codec crashing bug? should I be upgrading and testing?
17:58:40 Join criznach [0] (n=chatzill@host-69-145-134-192.grf-mt.client.bresnan.net)
17:58:46preglowtest_codec crashes again?
18:00
18:01:00 Quit ivan` (Read error: 104 (Connection reset by peer))
18:01:49 Join ivan` [0] (n=ivan`@unaffiliated/ivan/x-000001)
18:02:02Davide-NYCit has always crashed. It'll crash after a certain number of files are tested using the speed test folder option.
18:04:05Davide-NYC FS #7954
18:04:25preglowright, never tried that
18:05:36 Quit Gnu47 (Remote closed the connection)
18:05:49 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
18:15:57 Quit Gnu47 (Remote closed the connection)
18:16:09 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
18:16:12 Join MethoS- [0] (n=clemens@pD955C296.dip.t-dialin.net)
18:20:05n1shmm, I can't spot any obvious places to use the emac in midiplayer since the loop change...
18:21:17 Quit obo ("bye")
18:21:44preglowi don't expect emac will play a big role in midiplayer
18:22:03preglowit's just not the right application for it
18:22:13n1spreglow: so what do you expect will play a big role for midiplayer?
18:24:43 Quit nicktastic ("Leaving")
18:25:24 Join nicktastic [0] (n=nick@unaffiliated/nicktastic)
18:26:20 Quit Gnu47 (Remote closed the connection)
18:26:33 Join Gnu47 [0] (i=Gnu47@private.ntwk.thita.net)
18:26:43Nico_Ppondlife: seems to be working nicely
18:26:54 Join ilgufo [0] (n=matteo@host109-178-dynamic.5-87-r.retail.telecomitalia.it)
18:27:47 Nick parafin|away is now known as parafin (i=parafin@paraf.in)
18:27:51pondlifeyEP
18:27:54pondlifeOops
18:27:56pondlifeYep, even
18:28:06Mode"#rockbox +o scorche|w " by ChanServ (ChanServ@services.)
18:28:10Mode"#rockbox +b *!*Gnu47@private.ntwk.thita.net " by scorche|w (n=8dc5049d@rockbox/administrator/scorche)
18:28:17pondlifeNot had a rebuffering failure since.
18:28:20Nico_Pjust one thing... to calcuate the useful data amount, I need to know which handle is currently used by the codec
18:28:23scorche|wjust for a little bit ;)
18:28:38preglown1s: *shrug*, other optimization methods
18:28:56preglown1s: emac is good for filtering and bulk mixing
18:29:01pondlifeNico_P: What use is the "useful" amount? Just debugging?
18:29:01n1sheh,
18:29:12preglowmidi does mixing, but that is far from the bottleneck
18:29:23Nico_Ppondlife: sadly no, it's used to know whether the buffer is low
18:29:42Kick(#rockbox Gnu47 :message me when you arent disconnecting so much) by scorche|w!n=8dc5049d@rockbox/administrator/scorche
18:29:42pondlifeAh, ok... You can guess where I was going.. :)
18:29:54preglowso even if you did use emac for mixing in the old midiplayer code, i doubt it would have any performance effect nearing what the loop rewrite did
18:30:05Nico_Pbecause the handles are kept in memory, we assume the ones before the one being decoded are useless
18:30:09preglowi suspect it wouldn't even be noticable
18:30:56pondlifeNico_P: I assume they also need to have been bufclosed, right?
18:31:11pondlifei.e. bufclosed, but not yet overwritten?
18:31:14Nico_Ppondlife: precisely no, they are buclosed as late as possible
18:31:19pondlifeAh, ok
18:31:32 Part pondlife ("Gone")
18:31:38Nico_Pthe playback code keeps them open as long as possible
18:31:52Nico_Ppondlife: how safe is it to have a static var in the buffering code and give the playback code an accessor to set it?
18:32:37Nico_Pmaybe the accessor could be a queue event to make things safe
18:36:07 Quit desowin (Remote closed the connection)
18:37:40 Join BigBambi [0] (n=alex@rockbox/staff/BigBambi)
18:38:53 Join pondlife [0] (n=Steve@rockbox/developer/pondlife)
18:39:18Nico_Ppondlife: the teddybear effect worked, I did it via a queue event and I'm testing now
18:39:34pondlifeSorry, lost my internet
18:39:51Nico_PI talked without even noticing you were gone ;)
18:39:56pondlifelol
18:40:06pondlifeWhat was the change?
18:40:40pondlifeGuess I'll check gut
18:40:49Nico_PI added a static var in buffering.c telling which handle is currently decoded
18:41:11Nico_Pthe playback thread sets it via an accessor that sends an event to the buffering queue
18:41:26pondlifeAh yes, that should definitely be internal to buffering
18:41:56Nico_Psure but only the user knows, so a bit of a twist is required
18:42:25 Quit BHSPitMonkey (Remote closed the connection)
18:43:05*preglow hates playing register allocator
18:43:35 Join A_M [0] (n=anon3141@fw.dormnet.his.se)
18:43:48preglowamiconn: i hate it when i can't find anything to fill in the emac stall time with :/
18:43:53 Join lazka [0] (n=lazka@85.127.153.108)
18:44:51 Join bertrik [0] (n=Bertrik_@031-020-045-062.dynamic.caiway.nl)
18:50:46mrkikoey guys: the iRiver crash seems not to happen with voice menus disabled
18:50:50 Quit mrkiko ("leaving")
18:53:17 Join spiorf_ [0] (n=spiorf@host6-210-dynamic.14-87-r.retail.telecomitalia.it)
18:54:11Nico_Ppondlife: I'll commit and push very soon. it seems to work fine
18:55:08pondlifeOK, will carry on testing
18:56:08Nico_Pdone
18:56:20Nico_PI think "filling" is useless now
18:58:09*amiconn found a way to further speed up the ape filters
18:58:51amiconnFound that while trying a similar method as the one I impelented for coldfire on arm: that method gave a data abort 'cause the filter pointers aren't aligned
18:59:21amiconnThat means they aren't aligned on coldfire as well, which doesn't crash but performance suffers
18:59:28preglowamiconn: have you used mac.w or msac.w? it seems either gas or objdump has a bug
18:59:39amiconnwhere?
18:59:42 Part ataxic
19:00
19:00:05preglowamiconn: msac.w with load seems to be bugged in either objdump or gas
19:00:25*preglow finds cfprm.pdf
19:00:28amiconnI doubt that. Do you have an example?
19:00:46amiconnI mean your source vs. what objdump outputs
19:01:45preglow"msac.w %%a0l, %%d1l, (%[den])+, %%a0, %%acc1;" -> msacw %a6l,%a1l,%fp@+,%a1,%acc
19:02:05preglowcheck the first two regs there
19:02:37preglowi'm decoding by hand now to see if objdump messes up
19:02:40 Quit spiorf_ (Remote closed the connection)
19:02:40 Quit spiorf (Read error: 101 (Network is unreachable))
19:03:22amiconnThe only coldfire related bug I'm aware of in objdump is that it doesn't know move.w #IMM, %sr
19:03:46 Join webguest02 [0] (i=410848ee@gateway/web/cgi-irc/labb.contactor.se/x-854af2a533b1ae28)
19:04:36 Join spiorf [0] (n=spiorf@host6-210-dynamic.14-87-r.retail.telecomitalia.it)
19:04:50preglowobjdump bug
19:05:00amiconnBut what easily happens is that if you forget one % sign when using asm-in-C, gcc+gas mangle it into something different without warning
19:05:12amiconnHad that with the ape filters...
19:06:36 Quit Davide-NYC ("ChatZilla 0.9.78.1 [Firefox 2.0.0.8/2007100816]")
19:06:48 Quit J3TC- (Read error: 110 (Connection timed out))
19:10:22 Join webguest61 [0] (i=4fb36408@gateway/web/cgi-irc/labb.contactor.se/x-254429942a891cab)
19:10:30webguest61help please
19:11:12krazykitwe're not psychic, you have to ask a question before you can get helped
19:11:30webguest61ohh right :)
19:11:41webguest61my sansa san disk stuck
19:11:46webguest61when
19:11:51preglowok, i'm thorougly confused now
19:11:53webguest61i triend to open a game
19:12:01webguest61*tried
19:12:03krazykitwebguest61, please, the enter key is not the spacebar
19:12:22webguest61explain please?
19:12:38bertrikpressing the power button for 15 seconds or more should do a hard shutdown of the sansa
19:13:00webguest61thanks
19:13:04krazykitwebguest61, i was just asking you to ask the question on a single line instead of multiple lines. it makes it easier to comprehend.
19:13:05webguest61u helped me alot
19:13:14webguest61ohhh sorry krazykit
19:13:42webguest61thanks again betrick
19:14:10krazykitwebguest61, if this is one of rockbox's games, if you can reproduce the problem and describe how to do it, you could open a bug, if you're using the current build
19:14:57***Saving seen data "./dancer.seen"
19:14:58preglowamiconn: it looks like gas encodes the two multiplicands in mac/msac in the wrong order, lucky us multiplication is commutative ;)
19:15:12webguest61yea it's some game bug ,,, the game called mazze or mazam
19:15:50webguest61how i can store my TXT files in regular sansa sandisk firmware?
19:16:05krazykitwebguest61, you just copy them over like any other file
19:16:16 Join mf0102 [0] (n=michi@85.127.180.92)
19:16:48webguest61does sansa supports reading txt files?
19:17:04n1sin rockbox it does
19:17:23pondlifeNico_P: Did you see the MEM > 8 checks in playback.c... They might need moving into buffering.c?
19:17:42webguest61yeah i know but i want to ungrade my firm ver
19:17:52 Quit webguest61 ("CGI:IRC")
19:18:00pondlifeNico_P: Basically it only rebuffers when needed, not on ATA callback
19:18:06Nico_Ppondlife: they have been moved
19:18:09pondlifeCool
19:18:22Nico_Pit will also rebuffer on ata callback
19:19:02pondlifeI'm building now...
19:19:15Nico_PI haven't seen any problems yet
19:19:33Nico_Pand IMO it's much much cleaner
19:19:54pondlifeI was hoping that nearly all of the "filling" code could vanish
19:20:03Nico_PI'm hoping that too. looking
19:21:38preglowamiconn: this is starting to look like a gas bug
19:22:16 Join hcs [0] (n=agashlin@rockbox/contributor/hcs)
19:22:49preglowyeah
19:22:51preglowalmost certainly
19:24:01preglowusing msac, i get hangs, using the exact same code, but with neg + mac, no hangs
19:24:41 Quit barrywardell (Remote closed the connection)
19:25:52 Quit webguest02 ("CGI:IRC (EOF)")
19:27:15Nico_Ppreglow: two new commits
19:27:57Nico_Por pondlife maybe
19:28:28pondlifeTwo more!?
19:28:34pondlifeI only just built...
19:28:57Nico_Pheh, you're going to have another one
19:29:06pondlifeAlready on it... good work there again
19:30:51pondlifeDid you resolve that USB problem you mentioned?
19:31:01Nico_Pnot yet
19:31:36pondlifeOK. I've not seen problems there
19:31:40Nico_Pthe problem is that with the buclose calls now being queue events, they get to the buffering queue after the usb event
19:31:56Nico_Pbut the audio thread thinks they succeeded, somehow
19:32:31Nico_Pso maybe I should just remove all the handles after the usb connection, but that feels a bit extreme and wrong
19:33:16preglowarghg
19:33:32Nico_Ppondlife: I hope you're enjoying the perfect automatic track skip transitions :)
19:34:49pondlifePerfect? What improved?
19:35:21Nico_Psince earlier, not much, but I still find them nice to look at :)
19:35:22pondlifeAh, I've just had the little-bit-of-silence at the start of playback issue.
19:35:28Nico_Pdamn
19:35:50pondlifeI'd guess buffering needs to yield more?
19:36:00Nico_Pmaybe there should be a yield or short sleep after loading each track in audio_fill_file_buffer
19:36:12 Join obo [0] (n=obo@rockbox/developer/obo)
19:36:21Nico_PI think buffering yields enough
19:36:24pondlifeAh, ok
19:36:45pondlifeIs the audio thread waiting for buffering at that point?
19:36:55pondlifeI'd have thought not, as playback has started
19:37:01Nico_Pwould you mind trying to add a sleep(1) in the "do ... while" loop in audio_fill_file_buffer? to see if it solves it
19:37:26Nico_Pno but maybe the audio thread is keeping the codec thread from running
19:37:38 Quit spiorf (Read error: 110 (Connection timed out))
19:37:51pondlifeOK, it's reproducible, will try that
19:37:59 Join spiorf [0] (n=spiorf@host178-222-dynamic.8-87-r.retail.telecomitalia.it)
19:39:37Nico_Pthanks
19:40:16pondlifeYep, seems better
19:40:33pondlifeI put a sleep(1) at the end of the loop
19:40:34Nico_Pyield or sleep?
19:40:36Nico_Pok
19:40:42 Join stevenm [0] (n=stevenm@129-2-175-80.wireless.umd.edu)
19:41:06stevenmn1s, log stalking mentions a midi data abort?
19:41:09pondlifeWooh, audio_load_track() is only called in on place now
19:41:27pondlifeYou could put that inline and scrap continue_buffering?
19:41:45Nico_Pwhat do you mean ?
19:41:46n1sstevenm: yep, pixelma reported it but it seems to be fixed after my commit yesterday
19:41:54stevenmn1s, aah, yay.
19:42:16amiconnpreglow: Hmmmm, this is with 2.16.1?
19:42:19pondlifeNico_P: The routine audio_load_track() could be scrapped, and it's contents put into audio_fill_file_buffer()
19:42:35pondlifeSame for audio_initialize_buffer_fill()
19:42:39amiconnBecause, I didn't have msac.w problems in the libmpeg2 idct
19:43:03stevenmn1s, where do we go from here, besides that numberOfSamples thing? Increase number of voices on coldfire, or the sample rate?
19:43:10Nico_Ppondlife: I agree for the latter, but audio_load_track is quite long... I prefer to have audio_fill_file_buffer compact as it is
19:43:45stevenmn1s, I found it is easy to see if a file will play right on 44100: plug in USB during midi playback and it will set the sample rate to 44100. I don't know how, but that's what happens.
19:44:01n1sstevenm: I think going 44.1kHz would be nice but dunno if we're running fast enough...
19:44:04stevenmAt the moment, I think all 24 voices active cannot run at 44100 in real time. Just barely under
19:44:39stevenmAnd if this thing is to become a codec, it would need to be fast enough to handler other tasks at the same time, like UI. 44100 sounds better but not really all THAT much.
19:45:22pondlifeNico_P: Hmm, I just went into View Playlist and the marked track was way out..
19:45:44Nico_Ppondlife: after doing what?
19:45:45n1sbut if it becomes a codec the output will have to be resampled if it's not 44kHz...
19:46:14pondlifeNot much... but stop/resume went back to the wrong track too (the same one shown in Playlist Viewer)
19:46:22 Quit sslashes (Client Quit)
19:46:23pondlifeI may have tried to rewind at the end of track
19:46:27stevenmn1s, I was at one point thinking about trying to figure out seeking. Forward is easy enough, but backwards is more complicated.. haven't been able to think of anything other than rewinding the tracks and running them from the start (ticks only, not voices)
19:47:09stevenmto get a constant number of seconds, it would have to take tempo changes into account and all. when I am done with GRE and this other exam this week, I can try to look into it
19:47:09 Quit scorche|w ("CGI:IRC (EOF)")
19:47:19pondlifeThe playlist and resume point were 4 tracks earlier!
19:47:46 Join sslashes [0] (n=sslashes@209.67.252.122)
19:47:53Nico_Ppondlife: have you seen my latest commit?
19:48:02stevenmas for resampling, something like linear interpolation (or other methods) is a lot faster than rendering at twice the sample rate
19:48:04Nico_P"Forgot this one :("
19:48:36pondlifeAh, that might be it
19:49:09 Join scorche|w [0] (n=8dc5049d@st.iptel.by)
19:49:13Nico_Pthat's most probably it. on each auto track skip it would decrement the playlist position instead of incrementing it
19:49:27Nico_Pdumb mistake on my part
19:49:59pondlifeThat would explain my symptoms perfectly.
19:50:10 Join Llorean [0] (n=llorean@cpe-70-113-103-34.austin.res.rr.com)
19:50:13Nico_Psorry for the high commit rate
19:50:32pondlifeNo problem, they are (mostly) good things
19:50:34pondlife:)
19:50:41 Join stripwax [0] (n=Miranda@i-83-67-214-206.freedom2surf.net)
19:51:00 Quit stripwax (Client Quit)
19:51:10pondlifeDo you know if anyone else is testing much? GodEater_ , JdGordon, anyone else..?
19:51:34Nico_Pn1s: has been doing a bit of testing, but not fulltime I think
19:51:44 Join Crash91 [0] (n=evil91@196.218.80.170)
19:51:53Crash91hi
19:51:58pondlifeDo you know if any Archos users have tried it, just in case?
19:52:07Nico_Pno
19:52:18pondlifeI would see if you can round one up ;)
19:52:39pondlifeThey shouldn't notice any change at all, but worth checking before commit
19:52:45Nico_Pindeed
19:52:48egolosthmm.. how do i save a filename/path in rockbox?
19:53:48 Join mrkiko [0] (n=mrkiko@adsl-ull-185-119.42-151.net24.it)
19:54:08egolostis it a button on the virtual keyboard or something?
19:54:26stevenmn1s, well, battery is running out. I will be back later. Bye!
19:54:27pondlifeNico_P: Ah, I need to add my sleep(1) again?
19:54:28 Quit stevenm ("Connection reset by beer")
19:54:49Nico_Ppondlife: I added it, so no but you might get a conflict
19:55:07amiconnpondlife: That won't easily be possible
19:55:19amiconnAtm the hwcodec and swcodec playback engines are disjuct
19:55:23amiconn*disjunct
19:55:47pondlifeIndeed, I just wanted to ensure there had been no negative impact on HWCODEC
19:55:53amiconnAnd I don't think it makes sense to start converting archos playback now
19:56:17pondlifeNo, I wasn't suggesting any HWCODEC changes
19:56:18amiconnIt will once MoB is in, and proven stable and compact enough
19:56:30pondlifeJust to ensure that nothing had been broken
19:56:47 Quit hcs ("Leaving.")
19:57:05Nico_PI've just made sure it compiles
19:57:35Nico_Pand of course none of the buffering code is compiled
19:58:27pondlifeRewind at track end still gets confused btw, but no worse than SVN now
19:58:52Nico_PI fear there's no trivial fix for this
19:59:25pondlifeMust be detectable
19:59:34Nico_Pouch I have nearly 1k added to the gigabeat binary
19:59:39pondlifeOoer
19:59:49 Join hcs [0] (n=agashlin@rockbox/contributor/hcs)
19:59:58pondlifeHow did that happen? Debugging?
20:00
20:00:23Nico_Pyeah the debugging menu code probably adds a bit
20:00:34 Quit japc (Read error: 104 (Connection reset by peer))
20:00:47pondlifeAh, something else for someone to test - cuesheets
20:00:51Nico_Pand apparently the removals in playback.c don't compensate the addition of buffering.c
20:01:13pondlifeI'd expect a modest increase for the separation, but not much
20:01:20n1sstevenm: (for the logs) 22kHz upsampled to 44 will sound worse than just playing at 22kHz so IMHO 44kHz is a worthy goal :-)
20:01:29pixelmaegolost: on the c200 currently you accept the entered name/path with one of the both volume buttons in the virtual keyboard (accept and exit). I hope to improve the keymap soon (and if that happens update the manual) but I got two versions both have their upsides and downsides and am a bit undecided
20:01:44 Quit hcs (Client Quit)
20:02:49pondlifeNico_P: The rebuffer parameter for audio_fill_file_buffer() and audio_load_track() can be scrapped
20:03:00pondlifeIt's not used any more
20:03:05pixelmaegolost: with the power button you can exit without saving the changes
20:03:09 Join Frazz [0] (n=Fraser@thelawsons.plus.com)
20:03:45Nico_Ppondlife: very true
20:05:21pondlifeIs the audio_flush_and_reload_tracks() stuff still needed?
20:05:58Nico_PI'm not sure
20:06:22pondlifeWe seem to have multiple mechanisms for reloading after a playlist change.
20:06:41pondlifei.e. methods to invalidate the track list
20:08:00pondlifeaudio_invalidate_tracks() vs. audio_new_playlist()
20:08:01 Join TMM [0] (n=hp@c5147518c.cable.wanadoo.nl)
20:10:33preglowamiconn: yeap
20:10:56preglowamiconn: would newer binutils work out of the box for us?
20:11:07amiconnI don't know...
20:11:26 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
20:11:27preglowthere's something about new arch flags and all, but i don't know if they've kept the old ones
20:11:31Nico_Ppondlife: you're right, they seem a bit duplicate
20:14:06egolostpixelma: is there a stop button or paus only?
20:14:48 Join TotallyInfected [0] (n=ebola@pool-70-110-242-209.phil.east.verizon.net)
20:15:24 Join ompaul [0] (n=ompaul@gnewsense/friend/ompaul)
20:15:32 Join nomel [0] (i=4782332a@gateway/web/cgi-irc/labb.contactor.se/x-0f9cd5d323cef609)
20:15:48pondlifeNico_P: I'd suggest you change the parameter names of buffering_init(); the current ones are global vars too
20:15:55pixelmaegolost: for playing music?
20:16:39pondlifeplayback.c doesnt' need high_watermark any more
20:16:43egolostpixelma: found it.. long play.
20:17:53 Join Zagor [0] (n=bjst@46.35.227.87.static.tab.siw.siwnet.net)
20:18:13nomelwhere's the best way to report a rockbox bug for a device?
20:18:33nomelzagor: hey.
20:18:35pondlifenomel: Confirm it in the forums, then report on Flypray
20:18:38pondlifeFlyspray
20:18:43nomelk.
20:18:43BigBambiNico_P: I can do testing on H1x0 or gigabeat if you like/need it
20:19:02Zagorhi
20:19:15nomelzagor: i'm 90% sure it's a problem with the button code. i found that pressing pause then play again increases the volume and changes the radio station :)
20:19:21Nico_PBigBambi: testing on H100 would be useful, yeah
20:19:26BigBambiOK, will do
20:19:34nomelthat shouldn't have anything to do with initialization of the radio chip
20:19:38Zagornomel: !
20:19:42*BigBambi goes to look for the wiki page to find the repo
20:19:44nomelseriously!
20:19:47pixelmanomel: on what player?
20:19:51nomelc200
20:20:02nomelus w/radio version.
20:20:09pondlifeBigBambi: http://repo.or.cz/w/Rockbox.git?a=shortlog;h=mob might be useful
20:20:19BigBambipondlife: Cheers
20:20:35pixelmaI don't think it has anything to do with the initialisation just button action interference
20:20:51Zagornomel: confirmed
20:21:40Zagorpixelma: it's more than just a button bug. the frequency changes but rockbox doesn't think so
20:21:44nomel:-o
20:22:36Zagorhowever it's not the only radio bug. it still takes several seconds to just enter the radio screen for me. it was instant on my other c200.
20:23:09pixelmabut the volume change at the same time probably is ("play" is the same as "up" used for volume on some players, thought I kept an eye on that)
20:23:21Zagorpixelma: i agree
20:24:36nomeli'll be looking at the LO signal with a spectrum analyzer later today to see if there's any relation between what frequency the player is on and the actual.
20:25:07nomellunch time.
20:25:18pondlifeNico_P: Hmm, conf_preseek isn't used... might be the same in SVN
20:25:30Nico_Pno, it's used in svn
20:25:46pondlifeand conf_filechunk
20:25:50Nico_PI'm not sure whether it's useful to readd it
20:25:55amiconnpreglow: Btw, another objdump bug: trapf.w is disassembled as "sf %d2" (!)
20:26:01Nico_Pconf_filechunk is used in the buffering loop
20:26:14pondlifeAh yes
20:27:03pondlifeIt's not used in playback.c though
20:27:09amiconnZagor: How are the endpoints today? :)
20:27:50Nico_Ppondlife: it's used in codec_configure_callback
20:28:06pondlifeYes, it's set, but never read
20:28:29pondlifeIt's passed through perhaps to dsp_configure
20:28:37pondlifeAh, not even that
20:28:47Nico_Pyeah, you're right :)
20:29:03pondlifeconf_preseek too, right?
20:29:33Nico_Pyes, it's the same
20:30:04Nico_PI think I need to look into that callback. maybe I need to add something similar to buffering.c
20:30:21pondlifeFind out if it's really needed first
20:30:36pondlifeCodecs should be irrelevant to buffering.c
20:30:37preglowamiconn: marvelous
20:30:54pondlifeAny needed fudging should still be in playback.c, I'd think
20:31:02Zagoramiconn: very ... annoying.
20:31:18 Quit FOAD (Remote closed the connection)
20:31:37Nico_Ppondlife: the codec still needs a way to set these though
20:32:09pondlifeYes, but buffering.c should definitely not be involved
20:33:06 Join FOAD [0] (n=dok@dinah.blub.net)
20:33:35pondlifeMaybe you'll have to add a per-handle filechunk though
20:33:51pondlifeA global value for a mix of codecs won't work
20:34:07pondlifeIdeally it can be dropped.
20:36:01 Join petur2 [0] (n=petur@rockbox/developer/petur)
20:36:17Zagoramiconn: in fact I'm starting to wonder if the reason for my trouble is that the usb module isn't the same as in the i.mx31
20:37:27 Quit jhMikeS ("Meow!")
20:39:26Nico_Ppondlife: I don't see how I can make these conf_* vars configurable by the codec without making them global or adding code to buffering.c to allow changing them
20:39:27*BigBambi will start testing as soon as he can persuade Ubuntu to automount his iriver as anything other than read only
20:39:38Nico_Ppondlife: and they can't stay in playback.c either
20:40:18 Quit Sedgewick (Connection timed out)
20:41:13preglowZagor: they might have tweaked it subtly
20:41:35preglowthis is what we have disassemblers for :V
20:41:57Zagorpreglow: there are also many variations of this module. for example the MPC8349 contains a similar module, but which supports fewer endpoints etc.
20:42:07 Quit ompaul (Client Quit)
20:43:40preglowZagor: well, in that case it's not veru likely it's the exact same one as in the imx31
20:44:14amiconnI would not be surprised anymore if portalplayer introduced some extra bugs... :>
20:44:26 Join linuxstb [0] (n=chatzill@copernic-sda.pck.nerim.net)
20:44:27Zagorpreglow: exactly. and it would explain some of my headaches
20:45:45preglowhell, if they can manage to mess up an _instruction_, then this is surely right in their usual ballpark
20:46:39amiconnI don't think they messed the instruction in the core
20:46:57preglowprobably an interface issue
20:46:58preglowbut still
20:47:03amiconnThe problem is that swp(b) must lock the bus between the read and the write
20:47:03preglowit speaks volumes about their testing
20:47:17amiconn...and most probably that locking is what is broken, in the bus controller
20:47:44 Join TotallyInfected_ [0] (n=ebola@pool-70-110-242-209.phil.east.verizon.net)
20:47:56preglowyes, they probably don't mess with what they get from arm, but like i said...
20:48:04preglowmessing up bus locking requires some kind of a genious
20:48:07 Join Rob2222 [0] (n=Miranda@p54B1671C.dip.t-dialin.net)
20:49:39 Join mirak [0] (n=mirak@m94.net81-66-75.noos.fr)
20:50:16 Join jhMikeS [0] (n=jethead7@rockbox/developer/jhMikeS)
20:51:35 Join Lear [0] (i=chatzill@rockbox/developer/lear)
20:54:33pondlifeNico_P: Sorry, off now. I'm not sure what those configuration parameters are needed for exactly. Do you know?
20:55:01Nico_Penough I think
20:55:13pondlifeSo they are definitely needed?
20:55:37pondlifeIs it to ensure that buffer wrap is handled correctly?
20:55:48pondlifei.e. that we always have enough data?
20:56:01Nico_Pyeah I think they are. no, it's not exaclty that
20:56:18Nico_Pconf_filechunk is the amount of data read per loop
20:56:48pondlifeHow is that relevant to the codec?
20:56:51Nico_Pconf_watermark is the low buffer watermark. it depends on the codec and audio file
20:57:25Nico_Pwell apparently the codec will expect to be able to set it
20:57:47Nico_Psame for preseek
20:57:50Zagorconfirmed. DCIVERSION is '1' on the sansa, but '0' on the mx31
20:58:06BigBambiNico_P: Turning cross fade off whilst playing, track stopped (expected), restarted briefly (second or two), then stopped again. Resume playback then started that track from the beginning again
20:58:21BigBambiNico_P: And exactly the same thing turning crossfade on again
20:58:34pondlifeI got a bit of white noise, about 5 mins later
20:58:40pondlifeAfter playing for a while ok
20:58:45BigBambiNico_P: Is there somewhere you would like me to note things rather than spamming you here?
20:59:07pondlifeBigBambi: http://www.rockbox.org/twiki/bin/view/Main/MetadataOnBufferTesting might help
20:59:12mirakwhat do IDE do you use ?
20:59:14Nico_Pwhat pondlife said :)
20:59:15BigBambiCheers
20:59:45pondlifeSo, those codec params, how do they affect data availability later?
21:00
21:00:13pondlifeI can only think that they might mean the codec expects to be able to read data in chunks of a minimum size...
21:00:43pondlifePerhaps find out which codecs set those parameters to wacky values ;)
21:00:59pondlifeThen ask the relevant codec guru
21:01:08pondlifeOr perhaps amiconn knows...?
21:03:11Nico_Ppondlife: rgrep "configure" apps/codecs
21:03:36mirakcrap eclipse doesn't work un ubuntu gutsy
21:03:44 Quit TotallyInfected (Read error: 110 (Connection timed out))
21:04:31*BigBambi takes idiot hat off
21:04:53Nico_Ppondlife: I agree the filechunk setting is probably a bit useless, but the watermark at least is important and the preseek might be needed for proper seeking
21:05:03BigBambiI was going to say switching between tracks with different codecs takes ages, like 5 seconds, then I remembered I'd just turned crossfade on to test it
21:05:15 Quit jhulst (No route to host)
21:05:23pondlifeAh, yes, but it's much slower than SVN
21:05:30BigBambiYes, it still is
21:05:31Nico_Ppondlife: and now these settings are used by the buffering thread, so it needs a way to make them changeable by the codec
21:05:32pondlifeSomething is still sub-optimal there
21:05:34BigBambiCouple of seconds
21:06:03pondlifeNico_P: I guess they need to be per handle at least..
21:06:16amiconnhrmph
21:06:20Nico_Ppondlife: yeah maybe
21:06:28Nico_PBigBambi: skipping?
21:07:03pondlifeSeems silly to have them as globals when buffering is time-independent of playback.
21:07:17BigBambiNicoP; yeah, skipping from one track to another with different codecs
21:07:28Nico_Pok, I'll look
21:07:29BigBambiSometimes quite quick, sometimes slow
21:07:37Nico_Ppondlife: conf_watermark shouldn't be per handle
21:07:44BigBambiAnd this is all in the first 3 or 4 tracks, so should be in the buffer
21:08:09*BigBambi is liking the Buffereing debug screen
21:08:47pondlifeNico_P: Why not, what if you have an MP3 and an SPC track buffering; they have different watermark values...
21:08:52bertrikwould it help if i mapped the I2C space of the AS3514 (read-only)? or did someone already do that?
21:08:55preglowZagor: and DCIVERSION is what?
21:08:57Nico_Ppondlife: meh, I don't know anymore. I need to think of something else. I'll go make supper for myself
21:09:07pondlifeGood idea, it's been a productive day
21:09:11pondlifeSpeak tomorrow, maybe
21:09:18Nico_Phope so :)
21:09:30 Quit criznach ("ChatZilla 0.9.78.1 [Firefox 2.0.0.8/2007100816]")
21:10:06Zagorpreglow: Device Controller Interface version. it's a hardcoded read-only identification register that never changes.
21:10:32preglowZagor: so ours is actually newer than in the imx31?
21:11:02 Part egolost
21:11:15BigBambiNico_P: Skipping from MPC to OGG, 1st and 2nd tracks in playlist, All buffers full playing first track, skip to next track, continues playing while pcm buffer empties, then playback stops for a short time, then it all kicks back in, pcm buffer refills, next track starts
21:11:30ZagorI don't know what the number actually means. all I know is that the imx31 specs say the register should be all zeros, and the mpc8349 docs say their module should return 0x01 in that register. as the sansa does.
21:11:57amiconnpondlife: There's an even more complicated case: if there are several *very* small tracks with different format in the buffer ...
21:12:22bertrikdid someone already map the values on the I2C of the AS3514?
21:12:27 Join webguest06 [0] (i=415fa842@gateway/web/cgi-irc/labb.contactor.se/x-46eb71ee1624f099)
21:12:37amiconn..so that e.g. the first track stays below the low watermark entirely, and the second one has a different watermark. What to use in such a case?
21:12:41 Join J3TC- [0] (n=jetc123@dhcp78-127.njit.edu)
21:13:24BigBambiNico_P: Interesting, I am watching the debug screen on the main unit, music is playing back fine, and the LCD remote has a splash saying Codec failure!
21:13:35webguest06hi
21:14:13amiconnBigBambi: The splash is just not erased, as the debug screen doesn't handle the remote
21:14:17 Join Tavnos [0] (n=tavnos@lju91-5-88-174-161-25.fbx.proxad.net)
21:14:24TavnosHi there
21:14:30webguest06will you make a sleep mode (like the apple firmwar, cause thats the only reason i'm staying on apple's firmware
21:14:33Zagoralso the sansa DCCPARAMS register contains 0x387, which means it's neither the same as in i.mx31 (should have 'f' last) or mpc8349 (should have '6' last)
21:14:38BigBambiamiconn: So there was an earlier codec failure I didn't notice?
21:14:50amiconnI would think so
21:14:52BigBambiOK
21:14:58***Saving seen data "./dancer.seen"
21:16:12webguest06they will?
21:16:23BigBambiwebguest06: Patience grasshopper
21:16:32BigBambiFirstly, which iPod
21:16:32amiconnZagor: Do you know the meaning of those values?
21:16:52pondlifeamiconn: Can we maybe chat about the codec parameters tomorrow? I have to go now, sadly.
21:16:54amiconnYou said that the mpc8349 has fewer endpoints than the i.mx31?
21:17:27 Part pondlife ("Gone")
21:17:28Zagoramiconn: yes, the lowest five bits in dccparams is the number of endpoints
21:17:32amiconnpondlife: I really don't know how this currently works. I just know that there is a low watermark, and it's different per codec
21:17:53webguest06well, i have an iPod video (5g,old school)
21:17:59BigBambiAnd seondly I don't think anyone is working on one. I believe amiconn is looking at suspend for the 1st and 2nd gens as they physically cannot be turned off, but other than that I don't think anyone is working on it. Therefore there won't be one until someone does work on it.
21:18:14webguest06oh ok
21:18:18webguest06:(
21:18:19 Quit webguest06 ("CGI:IRC")
21:18:29 Join ompaul [0] (n=ompaul@gnewsense/friend/ompaul)
21:19:28 Quit A_M ()
21:21:24 Quit TMM (Read error: 110 (Connection timed out))
21:21:47 Quit miepchen^schlaf (Read error: 110 (Connection timed out))
21:21:56 Join miepchen^schlaf [0] (n=hihi@p54BF44F7.dip.t-dialin.net)
21:23:31Zagorhmm, scratch that confirmation. the i.mx31 docs are a bit fuzzy what the registers are supposed to contain.
21:23:42amiconnMeh. The demac filters are nasty. Not only v1 and v2 can be unaligned in the vector operations, but they can also be unaligned wrt each other :(
21:24:00bertrikoh you have docs now, nice
21:24:20 Join Sedgewick [0] (n=10334AEF@host99-233-dynamic.15-87-r.retail.telecomitalia.it)
21:24:50Zagorbertrik: I had docs all along. but I'm starting to doubt they are the right ones.
21:24:52amiconnSo that means in order to always fetch aligned, there need to be 4 cases. The effect might not be worth the overhead of checking :(
21:25:04 Join Buschel [0] (n=AndreeBu@p54A3F90F.dip.t-dialin.net)
21:25:58J3TC-So which bmp file is the bootlogo for iaudio x5?
21:26:06 Quit linuxstb (Read error: 110 (Connection timed out))
21:26:22J3TC-rockboxlogo.160x50x16.bmp ?
21:28:02 Quit spiorf (Remote closed the connection)
21:28:33 Quit ilgufo ("So Long, and Thanks For All the Fish - http://gufo.wordpress.com")
21:30:52mirakamiconn: have you tried something on slice.c ?
21:30:58Zagoroh no...
21:30:59 Join ilgufo [0] (n=matteo@host109-178-dynamic.5-87-r.retail.telecomitalia.it)
21:32:35Zagor387 was decimal. in hex it's... 183. three endpoints.
21:32:57amiconnOh wow
21:33:11amiconnA bit...limited, eh?
21:33:15Zagorquite
21:33:28Zagorno wonder i couldn't get #5 and #6 working
21:33:46amiconnHow many endpoints do you need for ums?
21:33:56Zagorjust two, so that's fine (plus one for control)
21:34:18amiconnThe control endpoint is fixed, correct?
21:34:31*amiconn doesn't know much about usb, just a little bit
21:34:35Zagorbut I added test endpoints higher up, which obviously didn't work very well
21:34:51Zagoramiconn: yes ep0 is always control
21:35:14Zagorbut I think it is included in the three. size 0 means "device mode not supported"
21:35:15amiconnHmm, for compound devices we would need more endpoints?
21:36:11Zagorno there's something called "configurations", where the host can select what type of device it wants us to be. but I haven't read up on it much. bertrik, do you know more?
21:36:18bertrikamiconn: yes
21:36:37bertrikZagor: as far as i know, there is in practice only one configuration
21:36:51bertrikbut it is possible to have more interfaces
21:37:04bertrikan interface is a set of endpoints
21:37:35Zagorah, I mixed them up
21:37:36bertrikso for example, one interface consisting of 2 EPs for MSC and another 2 EPs for virtual COM
21:38:16bertrikI mean, two interface of two EPs each
21:38:54ZagorI obviously need to read up on this
21:39:37amiconnHmm, the USB stack for my Amiga USB card (not open source, unfortunately) shows quite some details about that stuff
21:40:18bertrikthere is a very nice introductory text called usb-in-a-nutshell, I'll look it up
21:40:32Zagorbertrik: thanks, I've read it already :)
21:41:09bertrikIf you read that, and chapter 9 of the USB spec, you know almost everything you need to know about USB
21:41:33bertrikhttp://www.beyondlogic.org/usbnutshell/usb1.htm
21:41:37Zagorwell I may have skimmed some sections...
21:42:06Zagorhttp://www.beyondlogic.org/usbnutshell/usb5.htm#ConfigurationDescriptors
21:42:08 Join japc [0] (n=japc@bl7-241-88.dsl.telepac.pt)
21:42:14bertrikthe usbnutshell document is also available somewhere as a pdf (in case you want to print it)
21:42:38Zagor"Each configuration could be powered in the same way and draw the same current, yet have different interface or endpoint combinations."
21:43:06 Quit FOAD ("I'll be back")
21:43:28Zagorsounds like it should be possible to have MSC and COM in two different configurations
21:44:51mrkikoCan playing with USB be a dangerous activity on some situations?
21:45:07bertrikI think it is very unusual to have more than one configuration, I vaguely remember that windows does not even support it
21:45:11Zagormrkiko: you mean aside from corrupting your disk?
21:45:33n1smrkiko: if you play with it in the wrong neighbourhood?
21:46:15Zagorbertrik: do you remember if does not support as in bluescreen or as in ignores it?
21:46:46bertrikZagor: i don't know. I can't find the link anymore.
21:47:13amiconnZagor: Very interesting: The Amiga USB stack shows the number of endpoints for each connected device.
21:47:16n1sZagor: couldn't we have a setting on the device that controls which configuration is presented to the host?
21:47:26bertrikCan't we just drop off the bus and reappear as another device with another device id?
21:47:29Zagorn1s: yes we can
21:47:46amiconnIf I connect a mass storage device with a hardware controller (e.g. Iriver H300 or Iomega ZIP USB), they use 3 endpoints
21:47:47Zagorbertrik: I guess that's an option
21:47:57amiconnBut if I connect a PP device, it only uses 2
21:48:02Zagoramiconn: yes, lsusb on linux also shows all descriptors
21:48:15amiconnNote that this is fullspeed, not highspeed
21:48:29mrkikoI meant playing with it in general.
21:48:30 Quit Crash91 ("Bye Bye!")
21:48:30 Join DGMurdockIII [0] (i=DGMurdoc@64-184-11-103.bb.hrtc.net)
21:48:34DGMurdockIIIhi
21:48:54mrkiko... can I accidentally burn a device for example ?
21:48:55Zagoramiconn: are those 3 control + interrupt + bulk endpoints?
21:49:05Zagormrkiko: not likely
21:49:13amiconnThat's not shown, only the total number
21:49:16Zagorok
21:49:18DGMurdockIIIwhen is the creative vision M firmware going to be out
21:49:47ZagorDGMurdockIII: is there anyone working on a port for that?
21:50:39 Join FOAD [0] (n=dok@dinah.blub.net)
21:50:43DGMurdockIIIit says in the Development page that its a new port
21:50:55alienbiker99there is no set time
21:51:08*bertrik is mapping the I2C space of the AS3514
21:51:34n1sDGMurdockIII: it will likely be available when (if) it works somewhat (for example boots the player)
21:52:07rasherDGMurdockIII: That could mean anything from "almost usable" to "only information has been gathered - no code has been written"
21:52:40 Quit mrkiko ("bye bye - thank all!")
21:53:07Zagorfrom the CreativeZVMPort page: "Status: To-Do: Figure out how to put our own firmware on."
21:53:18Zagorthat's rather early
21:53:56*amiconn found a simple way of checking whether a specific asm code path is taken on coldfire
21:53:58bertrikI2C data repeats every 0x40 bytes
21:54:11pixelmamaybe there's more info in the "new ports" section in the forums? Haven't looked but I think I read something there
21:54:26amiconnJust put 'illegal' in that path. If it fires an illegal instruction exception, it is reached...
21:55:04amiconnOne could also put line A or line F emulation instructions for that...
22:00
22:05:15 Quit Xerion (Read error: 104 (Connection reset by peer))
22:06:16 Quit ompaul (Client Quit)
22:09:03 Join bluebrother [0] (i=Mq34Y4EG@rockbox/staff/bluebrother)
22:10:41 Quit mf0102 ("Verlassend")
22:10:50 Join barrywardell [0] (n=barrywar@89-125-27-44.dhcp-ripwave.irishbroadband.ie)
22:11:22 Join Xerion [0] (i=xerion@cp198589-d.landg1.lb.home.nl)
22:12:48 Quit Lear ("ChatZilla 0.9.78.1 [Firefox 2.0.0.8/2007100816]")
22:13:06 Quit atsea-22 (Read error: 104 (Connection reset by peer))
22:14:41Buschelamiconn: thanks for converting my patch to an .S-file and do the missing optimizations :o) now i have 8.9fps @30MHz
22:17:35 Quit amiconn (Nick collision from services.)
22:17:41 Join amiconn [0] (n=jens@rockbox/developer/amiconn)
22:18:16barrywardellbertrik: have you got the AS3514 datasheet?
22:18:29bertrikno, don't tell there is one
22:19:42 Join atsea-22 [0] (i=atsea-@gateway/tor/x-594c7638856a2f47)
22:19:45barrywardellyeah, we have access to it
22:20:46bertrikso I justed wasted some time ...
22:24:10 Part igorr
22:26:04 Quit Buschel ()
22:26:40 Quit Toxicity999 ("Leaving")
22:27:02 Join Toxicity999 [0] (n=bryan@unaffiliated/Toxicity999)
22:27:06 Quit ilgufo ("So Long, and Thanks For All the Fish - http://gufo.wordpress.com")
22:29:44 Join p3tur [0] (n=petur@rockbox/developer/petur)
22:31:05 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
22:31:32 Quit petur (Nick collision from services.)
22:31:37 Nick p3tur is now known as petur (n=petur@rockbox/developer/petur)
22:32:06 Join petur3 [0] (n=petur@ip-212-239-214-166.dsl-static.scarlet.be)
22:32:10 Quit petur (Nick collision from services.)
22:32:12 Nick petur3 is now known as petur (n=petur@ip-212-239-214-166.dsl-static.scarlet.be)
22:32:26 Join p3tur [0] (n=petur@rockbox/developer/petur)
22:33:19 Quit petur (Nick collision from services.)
22:33:23 Nick p3tur is now known as petur (n=petur@rockbox/developer/petur)
22:33:37 Part DGMurdockIII
22:33:53 Join petur3 [0] (n=petur@ip-212-239-214-166.dsl-static.scarlet.be)
22:33:57 Quit petur (Nick collision from services.)
22:33:59 Nick petur3 is now known as petur (n=petur@ip-212-239-214-166.dsl-static.scarlet.be)
22:34:13 Join p3tur [0] (n=petur@rockbox/developer/petur)
22:34:36p3turpetur: who the hell are you that you managed to steal my nick?
22:34:57 Quit Sedgewick (Remote closed the connection)
22:35:09 Quit petur (Nick collision from services.)
22:35:13 Nick p3tur is now known as petur (n=petur@rockbox/developer/petur)
22:35:32 Quit nicktastic ("Leaving")
22:35:43 Join petur3 [0] (n=petur@ip-212-239-214-166.dsl-static.scarlet.be)
22:36:36peturany channel op around please?
22:36:41japclol
22:36:44peturscorche?
22:36:46scorche|w?
22:36:50scorche|wwhat?
22:36:56peturcan you ban petur3?
22:37:00 Quit japc (Remote closed the connection)
22:37:08scorche|whe will still be on the network
22:37:19peturhe stole my nick, I just managed to change my password
22:38:00scorche|wbut what would banning him accomplish?
22:38:00preglowhow can i know you are you :P
22:38:23peturask something?
22:38:31scorche|wbeer?
22:38:41*petur grabs a bottle to kick petur3
22:38:47 Quit petur2 (Read error: 110 (Connection timed out))
22:38:47peturempty of course
22:38:56Mode"#rockbox +o preglow " by ChanServ (ChanServ@services.)
22:39:00pixelmaespecially since it's also a .be provider.. ;)
22:39:01Kick(#rockbox petur3 :preglow) by preglow!n=thomj@rockbox/developer/preglow
22:39:01 Join petur3 [0] (n=petur@ip-212-239-214-166.dsl-static.scarlet.be)
22:39:06preglowright
22:39:27peturI'm on telenet anyway, not scarlet
22:39:27scorche|wi can ban him, i am just trying to see what that would accomplish as he would be able to do the same thing, as he is still on the network
22:39:34Mode"#rockbox +b *!*@ip-212-239-214-166.dsl-static.scarlet.be " by preglow (n=thomj@rockbox/developer/preglow)
22:39:36Kick(#rockbox petur3 :preglow) by preglow!n=thomj@rockbox/developer/preglow
22:39:48*petur sens preglow beer of his choice
22:39:55petur*sends
22:40:02*scorche|w shrugs...
22:40:16preglowsure, but at least he won't be here pissing petur off :>
22:40:34Mode"#rockbox -o preglow " by ChanServ (ChanServ@services.)
22:40:52*preglow blinks at chanserv
22:41:21*scorche|w blinks at preglow
22:41:27peturI have seen my nick logged in in the past but thought it was just somebody using it unregistered, and ghosted him. But now he ghosted me, so somehow got hold of my password :(
22:41:30*BigBambi blinks at /me
22:41:45Zagorpetur: nasty
22:41:48*petur makes note of using stronger passwords
22:41:48rasherpetur: take it up with freenode staff
22:42:03rasheror is it resolved now?
22:42:03scorche|wpetur: well, irc does send passwords over plaintext
22:42:23rasherscorche|w: not if you use ssl connections
22:42:28scorche|wbut yes...if you have any network issues like that, join #freenode, and see what they can do (typically not much)
22:42:31preglowi keep sending my password to this guy called "nickserv" on efnet all the time :P
22:42:39scorche|wrasher: freenode doesnt accept ssl
22:42:59rasherscorche|w: ah, forgot. Why on the earth not
22:43:25pixelmaZagor: made an interesting observation re. radio and the "play" button on my c250... when the tuner is "confused" as I would call it, unpausing takes longer than when everythings alright
22:43:32scorche|wtoo complex...pretty much all ssl connections on other servers are flawed anyway...from what i understand, it is just hard to do with the prototcol
22:43:53scorche|won freenode, there is a tor/gpg combination
22:44:00preglowtor :/
22:44:16Bagdertor is... annoying at best
22:44:32scorche|wBagder: welcome back :)
22:44:32preglowtor would work well if people weren't so shitty
22:44:36preglowbut they are
22:44:44rasherNever had a problem with ssl on other networks. Anyway, freenode happily gave me my password when I forgot it once.
22:44:44*Bagder bows
22:44:53rasherWhich was somewhat scary
22:44:55preglowover half the people using it are trolls
22:45:02 Part TotallyInfected_
22:45:14Bagderand the non-trolls think tor is encryption ;-O
22:45:22scorche|wrasher: yes, but those implementations have large flaws...if you want to find out more, join #freenode and ask BearPerson
22:45:44 Join crashmatrix [0] (n=crashmat@s5590785f.adsl.wanadoo.nl)
22:45:47scorche|whe is one of the main ircd devs
22:45:59BagderI don't buy that
22:46:21*scorche|w shrugs...i am just repeating what i hear
22:46:31nomelpixelma, it doesn't change frequencies?
22:46:37crashmatrixevening, here with a quick question regarding my iPod health, is it okay to dd the entire disk, and reparition it, keeping firmware considerations in mind?
22:46:38nomelon my c200, it changes frequencies after unpausing.
22:46:47rasherFreenode also has a habit of making issues where there are none. But all this is off-topic anyway.
22:47:00nomelwell, it changes relative to the frequency that it's displaying.
22:47:14bbHi. I showed Rockbox on my Sansa e280 to a friend and he really liked it. Now he wants to know if it runs on his car stereo, but I have no clue. It's called "995DVD Car Entertainment System". Any idea were I could search for more detailed information what it really is?
22:47:15nomelnot sure if it changes when pausing also.
22:47:19 Quit DogBoy (Read error: 104 (Connection reset by peer))
22:47:31 Join donutman25 [0] (n=chatzill@65.75.87.48)
22:47:34crashmatrixbb: tough luck, probably not
22:47:39BigBambibb: Rockbox supported players are on the front page of the website
22:48:03BigBambibb: And that definitely isn't one
22:48:50bbBigBambi: it's a no name product and seems to be available under lots of different names and I guess I haven't found all names yet
22:49:02 Join Sedgewick [0] (n=Sedgewic@host99-233-dynamic.15-87-r.retail.telecomitalia.it)
22:49:09BigBambibb: Have a look at the front page
22:49:23pixelmanomel: not that I can tell (doesn't display a frequency change) and can't here it because frequencies are wide off and I only get static either way
22:49:36BigBambibb: I can can guarantee whatever it's names, rockbox does not run on any car radios.
22:49:40J3TC-Anyone playing with the sudoku patch?
22:49:49Zagorpixelma: it doesn't display a new frequency, it just hops
22:49:51J3TC-The one where you can pick the difficulty patch?
22:50:14 Join TotallyInfected [0] (n=ebola@pool-70-110-242-209.phil.east.verizon.net)
22:50:30pixelmanomel: ah... yes I could hear it now after a few tries it suddenly had a station
22:50:48pixelmabut it doesn't change volume anymore :)
22:51:38rasherJ3TC-: I believe that's one I wrote, but it was reported to have bugs. Why?
22:52:58J3TC-Ah awesome
22:52:59J3TC-Well
22:53:37J3TC-It crashes upon compiling and it gives the error of structure has no member 'menu_init' and 'menu_show'
22:54:01J3TC-So I'm assuming there is no more function for those two
22:54:31rasherAh, it'll need to be changed to use the same menu functions as the rest of the plugin, I guess
22:56:13J3TC-sudoku_menu_cb is also undeclared
22:56:14J3TC-Hrmm
22:56:28J3TC-Let me go look at the run through and see if I can make sense with this
22:57:53rasherIt should probably be a matter of mimicking the rest of the plugin
23:00
23:00:02J3TC-Hrmm, ok. Anyone know a command to just build the sudoku folder?
23:00:02J3TC-:3
23:00:30J3TC-This computer is slow and takes like 10mins or so just to get to compiling the sudoku
23:02:47 Nick parafin is now known as parafin|away (i=parafin@paraf.in)
23:04:26 Quit sounddude (Read error: 104 (Connection reset by peer))
23:04:58J3TC-Hrmm
23:04:59J3TC-That's odd
23:05:19J3TC-Menu_init is in the original sudoku.c
23:05:24n1sJ3TC-: if you do a "make" it will only build what's changed
23:05:26J3TC-I wonder why it was crashing
23:05:43rashern1s: given a slow enough computer and cygwin, that can still take ages
23:05:43 Join Isolinear [0] (n=A@c-76-105-254-119.hsd1.or.comcast.net)
23:05:49J3TC-Ah cool
23:06:56n1srasher: right, forgot it was _that_ slow, been years since I used it
23:07:33n1sJ3TC-: in which case "make rocks" should be slightly faster (will only make plugins and maybe codecs)
23:08:33 Join saratoga [0] (i=9803c6dd@gateway/web/cgi-irc/labb.contactor.se/x-9e668ceb15790785)
23:08:42J3TC-Hrmm
23:08:44J3TC-That's odd
23:08:48saratogalooks like sandisk is about to refresh the e200 series: http://audible.custhelp.com/cgi-bin/audible.cfg/php/enduser/std_adp.php?p_faqid=3648&p_created=1192116189&p_sid=XU6j3AOi&p_lva=&p_sp=cF9zcmNoPTEmcF9zb3J0X2J5PSZwX2dyaWRzb3J0PSZwX3Jvd19jbnQ9MiZwX3Byb2RzPTAmcF9jYXRzPSZwX3B2PSZwX2N2PSZwX3NlYXJjaF90eXBlPWFuc3dlcnMuc2VhcmNoX25sJnBfcGFnZT0xJnBfc2VhcmNoX3RleHQ9ZTIwMA**&p_li=&p_topview=1
23:08:59J3TC-Still crashes even though I did svn revert -R .
23:08:59J3TC-:3
23:09:20Zagorsaratoga: ah, so it's not a rumor
23:10:20n1sseems like a pure firmware upgrade though
23:10:22saratogai wonder if they're ditching PP or just updating the design
23:10:48saratogai don't think its just a firmware update, it'd be a little odd to stamp a new revision on the hardware if they're just patching the software
23:10:57saratogathough who knows
23:12:00preglowman, emac is so nice
23:12:00Bagder"I called Sansa and was told it will be a different player "
23:12:01n1sif they are still called e200 and the only major difference (to users) is support for another drm format my guess is on new firmware, they already did it with the rhapsody version...
23:12:08Bagder=> http://forums.sandisk.com/sansa/board/message?board.id=e200&thread.id=3477&view=by_date_ascending&page=2
23:13:17rasherWasn't it also said (rumoured) that the v2 would support sdhc?
23:13:27Bagderyes
23:13:29saratogahard to imagine it wouldn't
23:13:38saratogasandisk is selling a lot of SDHC stuff
23:13:50rashersaratoga: it's also hard to imagine why the firmware updates don't
23:13:55rasherfor the 1
23:14:06saratogathey want you to buy the 16GB v2 they'll release eventually
23:14:46rasherWell yes. I still wonder why they didn't launch with sdhc support
23:14:53saratogathat or maybe they're just not updating the firmware for the e200 anymore, if you prefer lazy over greedy
23:15:02***Saving seen data "./dancer.seen"
23:16:07saratogathe lack of a 16GB e290 is a bit odd anyway
23:16:17saratogathey 280 is so cheap now, you'd think they'd do something about that
23:16:41 Quit saratoga ("CGI:IRC")
23:19:57 Quit Tavnos ()
23:20:14 Join xregininflame [0] (i=48927114@gateway/web/cgi-irc/labb.contactor.se/x-9370de3a2b309f3b)
23:20:22xregininflamehey guys
23:20:39xregininflamei have a quick build question
23:21:09Bagderask away
23:21:17xregininflameok i just got subversion
23:21:22xregininflameim running it on windows xp
23:21:30xregininflameyes i read the wiki and manual and everything
23:21:35xregininflameand im confused
23:21:51xregininflamelike how do i get the trunk of rockbox?
23:21:53scorche|wwhat are you trying to do?
23:22:28xregininflameright now i just wanna checkout the source code
23:22:29scorche|ware you just trying to get the source? compile? patch?
23:22:32xregininflamebut i dont know how
23:23:02xregininflamejust trying to get the source im working on getting an enviroment thats c friendly right now
23:24:10scorche|wthe wiki suggests many ways of creating a dev environment...cygin, vmware, colinux, linux, etc
23:24:18xregininflameoh i know
23:24:24scorche|wif that is your end goal, you likely want to get that set up before getting the source
23:24:44 Join saratoga [0] (i=9803c6dd@gateway/web/cgi-irc/labb.contactor.se/x-b74e2a50b64286fe)
23:25:04xregininflamewell is there a way to set up subversion so like all i have 2 do is click something and i get the source?
23:25:11xregininflamei have hte .net framework right now
23:25:20scorche|wyou cant build rockbox with that
23:25:21xregininflamei dont like it too much but i can use it
23:25:28xregininflame.net? really
23:25:33xregininflameits c coding right?
23:25:38 Quit Arathis ("Bye, bye")
23:25:38scorche|wrockbox isnt .net...it is C
23:25:44 Join japc [0] (n=japc@bl7-241-88.dsl.telepac.pt)
23:25:59xregininflameright but the .net can compile c as well as c++
23:26:07scorche|wif you want to compile, etc on windows, you need cygwin, vmware, or colinux
23:26:09Zagorxregininflame: not for arm it can't
23:26:17xregininflameohh
23:26:19xregininflameok
23:26:21scorche|wor m68k, or sh
23:26:24 Quit jgarvey ("Leaving")
23:26:27xregininflamesee im horribly new at this
23:26:33xregininflamesorry...
23:26:45xregininflameill probably run colinux
23:27:01scorche|wany specific reason of choosing that?
23:27:35scorche|wi would recommend vmware, myself
23:27:36 Quit MethoS- (Read error: 101 (Network is unreachable))
23:27:40xregininflameumm i remebered the name
23:27:40rasherscorche|w: Why?
23:27:41xregininflamelol
23:27:47xregininflamevmware is better?
23:27:59scorche|wit is easier to set up than colinux
23:28:06 Quit jhulst (Read error: 110 (Connection timed out))
23:28:07xregininflamethats helpful
23:28:12scorche|wrasher: why what?
23:28:28peturor try virtualbox, it's lighter than vmware
23:28:44saratogahas NicoP said anything about the status of MOB? I'm wondering if he think it'll be committed to svn soon
23:28:48xregininflamei have to buy it??
23:28:54rasherscorche|w: why vmware over colinux? I'd imagine colinux being slightly faster, and perhaps slighly better integrated into Windows, whereas vmware is just a closed box
23:28:57scorche|wbuy what?
23:29:07xregininflamevmware
23:29:10xregininflameoh god
23:29:15xregininflamei didnt mean to start a debate....
23:29:17xregininflamelol
23:29:25saratogamaybe you should just read the instructions
23:29:25scorche|wrasher: just for the easy to set up factor, which seems to be a bit important
23:29:46scorche|wxregininflame: vmware player or vmware server are both free and will work
23:29:55scorche|wbut yes, please read the wiki page about it
23:29:59xregininflame..oh ok
23:30:11rasherscorche|w: how about the other points? I've never used colinux myself.. perhaps I should try some time
23:30:24xregininflameok so which is the better enviroment
23:30:33rasherxregininflame: go with vmware
23:30:48xregininflameok
23:30:57preglowsaratoga: it's getting there now
23:31:13scorche|wrasher: i havent used it myself either...i just know it is harder to set up, but is much easier on resource demands
23:31:21xregininflameshould i get the sever or player
23:31:35xregininflame<−− blushes cause he is a complete noob....
23:31:39rasherxregininflame: Player is simpler.
23:31:47scorche|wxregininflame: it doesnt matter...server has additional options which you probably wont need
23:31:57saratogapreglow: looking at the page theres only one bug outstanding, is that because its getting stable or because not many people have tested it?
23:32:07 Quit Sedgewick ("Money can't buy happiness but it can provide a better class of enemy.")
23:32:26preglowsaratoga: i wouldn't know, but i see nico_p working on it almost daily here now
23:32:52preglowsaratoga: plus, i saw some comments indicating it is pretty stable now
23:33:01xregininflameok im downloading the installer
23:33:27xregininflameok so about using subversion
23:33:31scorche|wyou dont need to update us :)
23:33:42scorche|wthe wiki tells you how to use svn
23:33:44Nico_Ppreglow: yeah, it's very colose from being committable :)
23:33:50xregininflamei thought u might like to know scorche :)
23:33:57scorche|wNico_P: dont jinx it ;)
23:34:39scorche|wxregininflame: http://www.rockbox.org/twiki/bin/view/Main/DocsIndex#For_Developers
23:34:42pixelmadon't remind scorche|w... ;)
23:34:49 Quit barrywardell (Read error: 110 (Connection timed out))
23:35:18saratogaNico_P: what do you need to do before it can be committed?
23:35:26preglowwhat samplerate does our voice files use, again?
23:35:54Nico_Psaratoga: there are a fex issues left, like an USB connection bug I need to find a good solution for
23:36:04xregininflameyea im still at a loss....
23:36:07Nico_Psaratoga: apart from that it's mostly testing and cleanup
23:36:18xregininflamedo i use if from the run command from windows?
23:36:39scorche|wno...you use it in the terminal of the vmware image when you get it set up
23:36:47xregininflameahh
23:37:01xregininflameso svn runs parrell with vm
23:37:09xregininflamewow speelling sucks
23:37:10xregininflamelol
23:37:18Nico_Psaratoga: you're welcome to have a look at the code ;)
23:37:24scorche|wno...it runs inside the vm....it is a program running inside the image
23:37:30 Join Sedgewick [0] (n=Sedgewic@host99-233-dynamic.15-87-r.retail.telecomitalia.it)
23:37:31xregininflameok
23:38:04*bluebrother wonders if it makes sense to /ignore lols
23:38:06xregininflamesorry im such an idot when it comes to this kind of stuff..ive never really dont it before :(
23:38:19scorche|wjust dont keep reminding us you are l;)
23:38:35xregininflame...love ya 2 man
23:38:36 Quit iamben (Read error: 104 (Connection reset by peer))
23:38:41bluebrotherscorche|w: but we might forget about it ... *g*
23:38:47xregininflametotally not in the sexual way....
23:39:04xregininflame:)
23:39:08saratogaugh stop posting
23:39:24xregininflamesorry..
23:39:29scorche|wxregininflame: have you, by chance, read the IRC guidelines linked in the topic?
23:39:56 Quit Frazz (Read error: 110 (Connection timed out))
23:40:16 Quit jhMikeS (Nick collision from services.)
23:40:22 Join jhMikeS [0] (n=jethead7@rockbox/developer/jhMikeS)
23:40:37xregininflameyes
23:40:37xregininflameoh
23:40:37xregininflamewhoops
23:40:37DBUGEnqueued KICK xregininflame
23:40:37xregininflameill be quite now
23:40:49CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
23:40:49*bluebrother is annoyed by his upcoming relocation
23:40:54preglowhrmph, big hunk of asm: 215% -> 225% for uwb file :///
23:41:45saratogawe really need to figure out what broke codec profiling
23:41:52saratogait would be useful for this sort of thing
23:42:44preglowso it is broken?
23:42:47xregininflamethanks guys
23:42:50xregininflamefor the help
23:43:54 Quit xregininflame ("CGI:IRC (EOF)")
23:44:18saratogapreglow: at least I can't get it to work and no one seems to have tried it recently
23:44:26saratogathe last time i did it was sept 2006
23:44:34preglowit never worked on arm, i think
23:44:45saratogait worked on the pp5002
23:44:52preglowhmm, ok
23:45:10saratogathough there perl script never worked so it was impossible to make sense of the results
23:45:31saratogai started fixing it a while back but gave up when i realized profiling itself wasn't working
23:45:44saratogathough maybe i'm just doing something wrong
23:45:55preglow884% -> 937% (nb), 401% -> 423% (wb). i guess at least it's worth commiting
23:48:35amiconnWhat target is that?
23:48:39preglowamiconn: coldfire
23:48:41n1ssaratoga: I ran a profiling build on my h300 maybe 2 months ago and it created output and all but the script seemed to have trouble so it couldn't match any symbols to the addresses which made the output less than usefull
23:48:42rasherWould nb be acceptable for voicefiles?
23:48:49preglowrasher: i would seriously want wb
23:49:04rasherStill slower than mp3 then
23:49:10rasheror at least not faster
23:49:24preglowthere are a couple of places here that will want some more asm
23:49:28preglowi think i'll hack qmf_filter next
23:49:40rasherHrm.. now I got duplicate entries in Database
23:49:41preglowrasher: these are pretty high quality files, though
23:50:02rasherpreglow: so are the files used on the codec testing page for mp3
23:50:19amiconnpreglow: Do you happen to know whether the penalty for unaligned memory accesses also applies to mac with parallel load from iram?
23:50:19preglowamiconn: i managed to fit the entire mem[] array in registers (evem for order 10), so i expected better performance than this
23:50:31amiconnI would think it does, but I'm not sure...
23:50:35preglowamiconn: well, i would certainly expect it to
23:50:58amiconnThat means it might be possible to get ape -c3000 usable on coldfire then
23:51:22saratogan1s: thanks, maybe its just me then, i'll try it again on arm and see if i can get output
23:51:25preglowi would be surprised if the parallel load was unaffected by alignment
23:52:03amiconnThat is, if the penalty for fetching word-word instead of longword is really 2 cycles, sclarproduct() takes almost twice as long as it could in 75% of all cases
23:52:29n1ssaratoga: the only other problem I had was that ggc ICEs on anything compiled with O3 with profiling enabled but that is probably a gcc 3.4.6 bug
23:54:10preglowamiconn: could you have a quick look and tell me if i'm doing something obviously stupid? http://www.pvv.org/~thomj/rockbox/filters_cf.S
23:54:12amiconnAnd I found that it's possible to fetch values into one of the registers currently processed in the mac instruction
23:54:23preglowamiconn: yeah, i've been doing that for ages :)
23:54:29preglowi even do it in this .S file, heh
23:54:38saratogahurray for pipelining
23:54:41amiconnThat allows me to give one register back for gcc to play with
23:55:02amiconnI can also give back one register in the vector add / sub functions
23:55:25saratogai want an arm target with EMAC or something similar
23:55:38preglowi THINK i can squeeze the den array into registers too, for .order_8, but not .order_10
23:55:38saratogabeating gcc in plain armv4 is harder then i like
23:55:53preglowsaratoga: yeah, it's slightly easier for coldfire
23:57:36 Join safetydan [0] (i=dc9d468b@rockbox/developer/safetydan)

Previous day | Next day