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

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

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

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

#rockbox log for 2010-04-01

00:00:06 Join Strife89 [0] (~michael@adsl-154-2-49.mcn.bellsouth.net)
00:02:17 Quit evilnick (Quit: Page closed)
00:03:26 Join webguest77 [0] (~5edf559a@giant.haxx.se)
00:04:10 Quit webguest77 (Client Quit)
00:07:13 Quit hd (Quit: Ω)
00:09:41 Join jd [0] (~jd@bas1-montreal33-1279639532.dsl.bell.ca)
00:09:41 Quit jd (Changing host)
00:09:41 Join jd [0] (~jd@Wikipedia/HellDragon)
00:15:41Blue_DudeI'd like to drop another keymap bomb in here. I noticed that there are optional key assignments for all sorts of things: playlist, track info, pitchscreen, and those are just for the WPS. I originally planned to replace the view playlist key with an assignable key but that's not going to be practical across all targets.
00:16:43Blue_DudeSo what about just leaving all the optional code in and just taking over a likely keypress? Would there be much heartburn if we lose the track info or pitchscreen keys but gain an assignable key?
00:16:53 Join komputes [0] (~komputes@ubuntu/member/komputes)
00:17:58gevaertsWhat do you mean? I don't understand "lose the pitchscreen keys"
00:20:23Blue_DudeWell, they'll still be in the code, but on some targets maybe they'll lose the dedicated keypress due to other keymap conflicts. It'll always be available in the context menu.
00:20:52 Join qfr [0] (void@unaffiliated/yw)
00:21:14qfrI am running Rockbox on a Sansa Fuze v1 and I was just showing the Doom game on it to my flatmate - doesn't it have any sound?
00:21:52LloreanOn many targets (can't say for yours) it has sound effects but no music.
00:21:52qfrThe normal playback of audio files works fine but the game doesn't appear to emit any sounds :(
00:21:58qfrHmmm
00:25:26 Join hd [0] (jd@bas1-montreal33-1279639532.dsl.bell.ca)
00:25:26 Quit hd (Changing host)
00:25:26 Join hd [0] (jd@Wikipedia/HellDragon)
00:26:46 Quit flydutch (Quit: /* empty */)
00:28:23 Quit jd (Ping timeout: 246 seconds)
00:28:53Torneaha, the proper bootcharting patch reveals that i'm an idiot and check_bootfile isn't even called on ipod
00:29:06Tornethe 1.5 seconds it takes when cabbie is seleected is indeed loading the wps
00:29:09Tornewhich isn't a huge surprise ;)
00:29:16Tornethe font still takes 6 seconds
00:29:23LloreanThat seems crazy
00:29:46LloreanWhat does it do for all that time?
00:29:49TorneLlorean: apparently it's always taken that long, or at least for a long time
00:29:59Tornei went back a couple of thousand revisions to before sbs/multifont/etc
00:30:08LloreanDid you compare 3.4 and 3.5?
00:30:13Torneand it still takes 6 seconds to load the font
00:30:20Tornenot yet
00:30:22 Join hebz0rl [0] (~hebz0rl@dslb-088-065-049-146.pools.arcor-ip.net)
00:30:36LloreanSomeone in a recent forum post complained that 3.5 booted slower than 3.4 on his... It was either an h100 or h300
00:30:41Torneyup
00:30:42LloreanI remember thinking "so it's not just ARM"
00:30:59LloreanDidn't expect it would be, or anything.
00:32:34linuxstbTorne: How far have you gone back?
00:33:07Tornebut yeah, with compeltely default settings, having already boted once so glyphcache exists, boot goes like this: font_load starts at 0.56 seconds, then finishes at 6.66 seconds
00:33:21Torneloading icons takes 150ms
00:33:41 Quit geertvdijk (Remote host closed the connection)
00:33:43Torneinit_tagcache takes 400ms even thugh it's not initialised and isn't laoding to ram
00:33:57Torneand then loading the wps takes 1.5s
00:34:01Tornetotal time 9.22s
00:34:23Torneso yeah. 2/3 of boot is font_load, 1/6 is wps loading
00:34:40Torne(to get to root_menu())
00:34:46gevaertsMaybe something related to glyphcache broke?
00:34:59Tornegevaerts: it's approx. equal time whether the glyph cache exists or not
00:35:17 Quit hd (Quit: Ω)
00:35:22gevaertsexactly. If it doesn't check properly anymore, this is what you'd get
00:35:28Torne..heh
00:35:45Luca_Suhm
00:35:48Luca_Sthe home page
00:35:52Luca_Shas been hacked again
00:35:56Luca_Sthis time, the word smurf
00:36:00Luca_Sis everywhere..
00:36:03 Join jd [0] (~jd@modemcable207.134-202-24.mc.videotron.ca)
00:36:04 Quit jd (Changing host)
00:36:04 Join jd [0] (~jd@Wikipedia/HellDragon)
00:36:29pixelmahahaha
00:36:39perfectdrugwodz: (for the logs) I started on the SVG for the mpio hd200 manual, i put it on flyspray later today for review
00:36:52Luca_SSmurfa Fuze it's an amazing name :D
00:37:15Luca_Soh well... it's the first of april already? :D
00:37:21linuxstbTorne: What's the earliest revision you tested?
00:37:51Tornelinuxstb: 23257
00:37:56Tornethe one immediately before SBS was committed
00:38:34Torneso, after 3.4
00:38:37 Quit Geert_van_Dijk (Quit: ChatZilla 0.9.86 [Firefox 3.6.2/20100316074819])
00:39:14linuxstbTorne: Hmm... font.c at least wasn't changed between 18609 (Sept 2008) and 23742
00:39:55Torneyeah, i noticed
00:40:07linuxstbI guess 18608 would be worth testing...
00:40:10Tornenothing significant seems to have changed about fonts, other than multifont :)
00:40:44Tornei'll try various points
00:41:05Torneit'll be slightly inconvenient to go back past the logf rework and keep bootchart though :)
00:41:30Tornehm, actually if it's just fonts i can do taht only
00:41:56 Join funman [0] (~fun@rockbox/developer/funman)
00:42:08pixelmadiacritic support was added not too long ago, wasn't it?
00:42:23linuxstbpixelma: Yes, but Torne has checked before that.
00:42:23Tornethat was after the revision i tested also, i think
00:42:36pixelmaok
00:42:41Tornei'll just go back a bunch and do a two line change to just time font_load
00:42:48Torneand see if i can find anything ;)
00:43:00qfrHaha I should write an NFO viewer for Rockbox
00:43:07funmanhello-smurf
00:43:10linuxstbTorne: Or did the font itself change?
00:43:19Tornelinuxstb: hm, possible i guess
00:43:32Tornelinuxstb: i'd like someoen to try on a different player entirely, also
00:43:49Torneand jsut see if the boot times are proportional
00:44:13*linuxstb looks around for a charged target...
00:45:53 Quit perfectdrug (Ping timeout: 246 seconds)
00:45:59*Torne grabs the bzr bisect plugin :)
00:46:03*pixelma can't see one joined
00:48:16 Join perfectdrug [0] (~marko@p5B0EC16F.dip.t-dialin.net)
00:48:38 Quit bertrik (Quit: De groeten)
00:49:44linuxstbTorne: Do you have a patch somewhere?
00:49:51funmanqfr: i have sound in doom on fuzev1, perhaps it depends of the WAD file ?
00:49:59 Quit jgarvey (Quit: Leaving)
00:50:06qfrI haven't installed any WAD file D:
00:50:12qfrJust the normal setup
00:50:14qfrFreeDoom
00:50:17qfrOr whatever it is called
00:50:54funmanlooks like you need to enable sound in the menu: it's off by default
00:53:28qfrOh.
00:54:21 Part captainkewlllll
00:58:59amiconnTorne: Font loading can't always have been that slow. Archos recorder (with an 11MHz CPU!) used to boot in 5..6 seconds
00:59:21Torneamiconn: yeah, i'm bisecting now
00:59:24amiconnThat's from power-on to menu (with rockbox in flash but obviously loading fonts and stuff from disk)
00:59:30Tornebut it's slow ;)
00:59:51amiconnI'm quite sure this slowdown happened less than half a year ago
01:00
01:00:35***Saving seen data "./dancer.seen"
01:00:48amiconnI don't use any bitmaps in my wps'es but I always use an on-disk font (usually rockfont on small displays and some nimbus variant on the larger ones)
01:01:00qfrAh funman, you were right! Fabulous!
01:01:08 Quit Stephen__ (Quit: Leaving)
01:01:20Torneamiconn: well it only takes 1.5s to load the cabbiev2 wps for me
01:01:32Torne6s for the font cabbiev2 uses :(
01:05:48Luca_Ssmurfy night to everyone!
01:05:51 Quit Luca_S (Quit: CGI:IRC)
01:07:33archivator...and rbutil now supports parallel voicing ... a good night for voice users, methinks :)
01:07:37 Quit archivator (Quit: Leaving)
01:07:45 Join Darkknight512 [0] (~Darkknigh@CPE00212968356c-CM00186845dd46.cpe.net.cable.rogers.com)
01:08:54 Quit dfkt (Read error: Connection reset by peer)
01:09:51 Quit ender` (Quit: A computer program will always do what you tell it to, and seldom what you want it to.)
01:10:07 Join crculver [0] (~crculver@59.178.196.106)
01:10:58crculverI reformatted my ipod's vfat file system after disk corruption. Although I can connect in disk mode and reinstall rockbox, the rb bootloader now tells me "No partition found"
01:11:31funmangood news: doom runs fine on Clip+ !
01:12:09pixelmafind the blue door?
01:12:11Schmoyeah
01:12:12Tornecrculver: what dd you frmat it with, and wich ipod is it?
01:12:19 Nick Schmo is now known as Schmogel (~Miranda@p3EE21EF1.dip0.t-ipconnect.de)
01:12:53crculverTorne, It's an iPod Video, 30gb, I used mkdosfs to format it.
01:13:10 Join smurf [0] (~58d9759a@gateway/web/freenode/x-dmlphevqxmueiugy)
01:13:10Torneyou need to explicitly specify the sector size
01:13:37Tornethe 5.5G disks are supposed to be formatted with 2048 byte sectors
01:13:56Torneadd -S 2048
01:14:04 Quit jd (Quit: Ω)
01:14:31crculverTorne, Thanks. Too bad I'll have to copy all my music all over again. :-\
01:16:12 Quit komputes (Ping timeout: 268 seconds)
01:16:31 Join Tomis [0] (~Tomis@70.134.100.153)
01:16:32Torneamiconn: 18086 also takes 6 seconds to load the font
01:17:18 Join jd [0] (~jd@modemcable207.134-202-24.mc.videotron.ca)
01:17:18 Quit jd (Changing host)
01:17:18 Join jd [0] (~jd@Wikipedia/HellDragon)
01:17:49crculverTorne, I continue to get the error.
01:18:37Tornecrculver: try with mformat from mtools
01:18:46Tornehttp://www.rockbox.org/wiki/IpodManualRestore
01:18:57Tornethis is how peopel ahve restored without itunes in the apst
01:19:11Torneeasier option, if you have any access to a windows machine: use itunes :)
01:19:51Torneamiconn: so tha's well past the last 6 months..
01:20:06Torneamiconn: that's the same 18 months ago
01:20:25Torne22 months even
01:21:50Torneamiconn: maybe other people are not having this issue :)
01:22:17Torneoh wait, i forgot about the glyphcache
01:23:20 Quit Llorean (Quit: Leaving.)
01:23:35 Join evilnick [0] (~48e51a1b@rockbox/staff/evilnick)
01:27:38 Part toffe82
01:28:25crculverThe iPod Video is 5.5G?
01:28:34Blue_DudeAll right, I'm ready to commit the hotkey patch. Any last words?
01:29:13Tornecrculver: some of them are
01:29:22 Join merbanan [0] (~banan@c-62-220-165-110.cust.bredband2.com)
01:29:28Tornecrculver: the 30GB model it's ahrd to tell without checking the serial number
01:29:46pixelmaBlue_Dude: your keymaps look a bit hmm... too simple
01:29:58 Join komputes [0] (~komputes@ubuntu/member/komputes)
01:30:02Blue_Dude?
01:30:03guymannlove the new site design
01:30:30Torneamiconn: yeah, i've gone righ back to 15756 and it still takes 6 seconds to load a font, and that was before cabbie was even *included*
01:30:51Blue_Dudepixelma: I didn't design the keymaps, just stole a key from another function.
01:30:55Torneamiconn: so it looks like font loading *has* always been slow on ipod video 5.5g
01:31:10pixelmaBlue_Dude: from what I remember about looking at it today in the morning. The playlist viewer shortcut also had the problem, I know
01:31:16Torneamiconn: i can't go back much further or 5.5g won't even boot :)
01:32:06Blue_Dudepixelma: Oh, I see. Yeah, I think all the keymaps need revisiting. A lot of the assignments are kind of random. DevCon topic, maybe?
01:32:13Torneso i think it's safe to say that whatever slowdown other people are experiencing i'm not reproducing it here 
01:33:43pixelmaBlue_Dude: I don't mean randomness, I mean "unclean" implementation - many just assign a button combo without checking pre conditions so that you need to be pressing the two buttons *exactly* at the same time, especially if one or both keys are already assigned tto other functions
01:33:48Blue_DudeAnyway, all I'm doing is stealing a key and the default is to duplicate the same function. The average user shouldn't notice any changes. I think we need to have the conversation about rethinking keymaps, but that's not what this patch is for.
01:34:25pixelmaand it's not so easy to improve the keymaps
01:34:51Blue_Dudepixelma: I entirely agree. Some top down rethinking is in order.
01:35:07Blue_DudeBut maybe not right this minute. :)
01:35:18pixelmaI'm not so positive that this would help any
01:35:41Blue_DudeNo it won't. But it doesn't hurt anything either, which is something I tried hard to avoid.
01:36:08pixelmaBlue_Dude: you didn't say something about the targets that don't have the playlist viewer shortcut at all... I don't expect tthat they'll get it, just making sure you are aware
01:36:14Blue_DudeMaybe it'll prompt some skull work though.
01:36:56pixelmaamiconn: does the Iaudio M3 also have retrictions in button combos?
01:37:01Blue_DudeYeah, I noticed, and those targets are switched off in their configs. They can be switched on after some keymap redesign.
01:37:04pixelmarestrictions too
01:38:49 Join kugel [0] (~kugel@rockbox/developer/kugel)
01:38:53Blue_Dudepixelma: I only switched on those targets that already had the VIEW_PLAYLIST key that also didn't conflict with any tree/list/standard keypresses. That left quite a few targets ready to go without further work, but a number need another look.
01:39:15kugelpixelma: "just assign a button combo without checking pre conditions so that you need to be pressing the two buttons *exactly* at the same time" I'm not sure if this is true
01:39:24Blue_Dudekugel: You magnificent bastard, I read your post!
01:39:34pixelmakugel: read on
01:39:53*Blue_Dude watched the intro to Patton recently.
01:40:33Blue_Dudekugel: Still not sure what you meant though...
01:40:48kugelwhat did I do? :(
01:40:56kugelpixelma: ?
01:42:00amiconnpixelma: yes
01:42:50pixelmaif the buttons (used on their own) already trigger an action, then you have to make sure that this one isn't already true before you can hold your button combo (hence pre conditions are needed)
01:42:52amiconnNo button combos possible except with PLAY. But long Play is shutdown so it's not really useful. Applies to both main and remote.
01:43:06pixelmathat was to kugel
01:43:11Blue_Dudekugel: I'm referring to your comment on FS #11081. I did go back and remove the conditionals around wps_view_playlist, so it will compile all the time. That's to open up remapping possibilities without forgetting something important.
01:43:57kugelwps_view_playlist?
01:44:15pixelmaI really really doubt you'll find more targets with a spare button... the keymaps are already quite crammed
01:44:17Blue_Dudepixelma: Yes, some keypresses need to have their pre-conditions explicitly defined, other than the all-too-common BUTTON_NONE.
01:44:45Blue_Dudepixelma: Which is why I resorted to stealing an existing one.
01:45:06kugelthe last patch has "case ACTION_WPS_VIEW_PLAYLIST:" #ifdef'd. I suggested to remove the entire case and move the code (and the #ifdef) into "case ACTION_WPS_HOTKEY:"
01:45:15Blue_DudeShameless but effective! Especially when also defining the same behavior as the default.
01:45:36amiconnTorne: Now that *is* odd. It also means the behaviour you're observing is not the same as on other targets :\
01:45:40pixelmawell even that one wasn't easy... and maybe doesn't work everywhere (see the statements about the M3)
01:46:17Blue_Dudekugel: I did take out the ifdefs, but I can't move the code to ACTION_WPS_HOTKEY. If I did, the key wouldn't be reassignable. You might not *always* want the hotkey to behave that way.
01:46:20kugeleffectively replacing the ACTION_WPS_PLAYLIST action with ACTION_WPS_HOTKEY, removing the need to put all the ifdefs in the target key files (since both actions are mutually exclusive)
01:46:20pixelmathe playlist viewer shortcut key is one of the most recent additions
01:46:37amiconnTorne: Is that an 80GB G5.5 (precisely one with an MK8011GAH)?
01:46:40Torneamiconn: yes
01:46:47kugelBlue_Dude: I'm looking at your latest patch, I'm not sure what you're looking at :p
01:46:48pixelmain keymaps (except new ports)
01:47:02Torneamiconn: er, no, 8010GAH
01:47:05Blue_Dudekugel: I didn't upload the very latest.
01:47:06kugelpixelma: that's not what you said
01:47:10*amiconn wonders whether the slow font loading might be due to the 1K sector handling
01:47:12Blue_DudeBut i will now.
01:47:22kugelBlue_Dude: well you should if you want that my talk makes sense :)
01:47:23pixelmakugel: what?
01:47:43Torneamiconn: also, re your comments eearlier about handling sector sizes better, i am aware of hte problems and i'm gonna experiment with it, i ahve this drive to test on.
01:47:51kugelyou said you need the precondition since otherwise the buttons all need to be pressed exactly at the same time
01:47:57 Quit perfectdrug (Ping timeout: 265 seconds)
01:48:08amiconnSeems like I need to test boot speed myself on various targets. I won't do that now though
01:48:08pixelmaand explained later why
01:48:21Torneamiconn: well, I'm going to post my bootcharting patch on FS in a sec
01:48:31Torneamiconn: i want to put it in svn but the macro usage may be objectionable
01:48:42pixelmakugel: I didn't go into detail that much though but it was still in the same sentence
01:48:47kugelpixelma: you said "especially" which sounds as an additional problem, not like an explaination
01:48:48pixelma+detail
01:49:14Torneamiconn: it would probably make this easier, anyway, because it's easier to compare results this way :)
01:50:41Blue_Dudekugel: OK, it's up.
01:51:41Blue_Dudekugel: the more I work on it, the less the patch does to the existing code. Maybe that's a good sign. :)
01:52:42kugelBlue_Dude: so, the playlist viewer is now accessed via onplay instead of directly returning VIEW_PLAYLIST from the wps?
01:53:04kugelanyway, I'm not sure why ACTION_WPS_VIEW_PLAYLIST needs to stay in the keymap files now
01:53:17Blue_Dudekugel: Yes, exactly. The hotkey execution doesn't go into the WPS code at all.
01:53:24Torneamiconn: linuxstb: gevaerts: http://www.rockbox.org/tracker/task/11161 <- bootchart patch, finalised
01:53:36Blue_Dudekugel: That's why I ifdef'd it out to begin with...
01:53:42Torneanyone who has an opinion on use of weird macros, please have a look :0
01:53:54Tornegoing to bed now tho.
01:54:32kugelBlue_Dude: what happens if I #undef HAVE_HOTKEY? can I still access the playlist?
01:55:33Blue_Dudekugel: You can always access it. But if HOTKEY is defined, it just won't have its own dedicated keypress.
01:55:46kugeljust a minor thing, I think it should have been HAVE_HOTKEY instead of just HOTKEY
01:56:04Blue_DudeYou can still get to it from the context menu for instance.
01:56:13Blue_DudeOr the main menu, etc.
01:56:33kugelI mean using the short cut that's in svn
01:56:44Blue_DudeHm. I can do a global search and replace on that. Not a bad idea.
01:56:56 Quit smurf (Quit: happy smurfing)
01:57:07Blue_Dudekugel: ???
01:57:29kugelsvn has a hotkey in svn that goes to the playlist viewer from the wps
01:58:01kugelit looks like you remove it for the !defined(HOTKEY) variant
01:58:20 Join perfectdrug [0] (~marko@p5B0EC16F.dip.t-dialin.net)
01:58:54kugel(I can imagine that instead of "(void)hotkey" "if (context == CONTEXT_WPS && hotkey) return ONPLAY_VIEW_PLAYLIST" would revive it
02:00
02:00:09Blue_Dudekugel: I see what you're saying. No, I didn't take anything out. Where are you looking?
02:00:17pixelmakugel: the now hotkey replaces the playlist viewer shortcut - with the default being a shortcut to the playlist viewer
02:00:22kugelBlue_Dude: still your patch
02:00:52kugelthe case ACTION_WPS_VIEW_PLAYLIST: is removed as far as I can see
02:01:22kugel(as I suggested, but without ability to do what it does now without HOTKEY defined)
02:02:13 Quit z35 (Ping timeout: 252 seconds)
02:02:29pixelmaah, now... you're speaking about the targets that have the playlist viewer shortcut but don't have the hotkey because of conflicts in lists
02:02:47kugelyep
02:03:14Blue_Dudekugel: I originally defined that case out because it wasn't needed anymore - assuming that that key would always have been replaced. Since that may not be true, I'm leaving it in for now.
02:03:33kugelwhy?
02:03:40Blue_Dudekugel: but if the keypress isn't replaced, HOTKEY isn't defined.
02:03:47Blue_DudeSo there's no change.
02:04:06kugelyou again don't understand what I'm saying
02:04:26Blue_DudeNo, I guess not... :(
02:05:01kugelACTION_WPS_HOTKEY basically replaces ACTION_WPS_VIEW_PLAYLIST, so the ACTION_WPS_HOTKEY case should do what the ACTION_WPS_VIEW_PLAYLIST case does now, for targets that don't #define HOTKEY
02:07:12kugelplus, because of that ACTION_WPS_VIEW_PLAYLIST isn't needed anymore in the keymap files
02:08:59Blue_Dudekugel: I can do that, but the key label won't be very indicative for non-hotkey targets. Besides, I might try to steal another keypress other than VIEW_PLAYLIST on those targets and leave VIEW_PLAYLIST alone.
02:09:27kugelbut they are mutually exclusive?
02:09:31Blue_DudeAdding the case won't hurt anything, but if the keymap is correctly setup, it won't ever fire either.
02:09:53Blue_DudeThat was the original plan, but it's not essential.
02:10:11 Quit perfectdrug (Ping timeout: 265 seconds)
02:10:21kugelwhy should a target have the playlist key and the hotkey?
02:10:37 Nick fxb is now known as fxb__ (~felixbrun@h1252615.stratoserver.net)
02:10:39kugelthat doesn't make sense since hotkey can do that, it is the default even
02:10:41crculverTorne, Alright, that was a long process, but in the end it worked, I have rb back. Thanks for your help
02:10:44Blue_DudeI could take over SHOW_TRACK_INFO or PITCHSCREEN or something else instead.
02:11:15Blue_DudeIn which case perhaps those could be the default behavior instead.
02:11:32kugelor make it so that track info and pitchscreen can be put on the hotkey instead :)
02:12:05 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
02:12:54kugelwhat you describe doesn't appear to be likely to happen, and we usually don't code for the future
02:14:14Blue_DudeWhich I did do...
02:14:38 Part crculver ("Leaving")
02:15:23kugelthe keymaps do have the hotkey and playlist key mutually exclusive (in your latest patch), and it's the same button for all targets. I don't see why to keep the playlist button
02:15:28Blue_DudeSo far, delete, view playlist, open with, show track info, pitchscreen, and insert playlist can all be assigned.
02:16:16Blue_DudeThey are only mutually exclusive for the targets that are ready to be switched on. They are not mutually exclusive by design. That was the original intention, but it's not strictly necessary.
02:17:05Blue_DudeI can remove the conditional ifdefs from the keymaps. I just thought it would be easier to troubleshoot if I didn't though.
02:18:26 Join CaptainKewl [0] (jds@207-237-106-60.c3-0.nyr-ubr1.nyr.ny.cable.rcn.com)
02:20:25Blue_DudeI just thought of a reason to keep the define as HOTKEY instead of HAVE_HOTKEY: maybe the define itself can designate which key is being replaced. It would solve some problems. Have it be: #define HOTKEY PLAYLIST or #define HOTKEY TRACK for instance.
02:21:13kugelCONFIG_HOTKEY then :)
02:21:44 Join z35 [0] (~z35@ool-18bd3f51.dyn.optonline.net)
02:22:23kugelI think I didn't add the playlist viewer to all targets because of lacking buttons, some of them had a track info hotkey; I guess it would make sense to use that button instead then
02:22:25Blue_DudeHm, I could see that.
02:22:40 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
02:22:42kugelbut I don't think the defaults should differ across targets just because of that
02:23:16Blue_DudeThe conditional defines would go back in, and the default logic in hotkey.h would be a little more complex, but so be it. It works better.
02:23:58kugelwhat conditional defines?
02:24:33Blue_Dudekugel: That's an education and manual issue. In making it transparent, you're going to get some compromises somewhere.
02:24:42Blue_DudeThe ones in wps.c
02:24:51Blue_DudeAround the case.
02:24:52kugelI don't understand
02:25:11Blue_DudeWhich? :)
02:25:49kugelI hope you're not trying to make the default conditional on what button you've stolen
02:26:39Blue_DudeWell, yeah, that would be transparent, wouldn't it? Someone would miss that special button and complain and then where would we be?
02:27:00kugeltell him to change the hotkey then
02:27:39kugelwe absolutely hate different defaults across targets, and "being transparent" is no justification
02:27:43Blue_DudeYah, I'm right there with you on that, but Joe User probably doesn't care. He wants it to work the same.
02:27:55kugelWe worked hard to make rockbox the same across so many targets
02:28:17Blue_DudeThen the default ought to be "no hotkey set" and let the chips fall where they may.
02:28:35kugelI don't see why
02:28:46Blue_DudeAnd then I can get rid of hotkey.h too.
02:29:42Blue_DudeThere's no point in preassigning a key function if it duplicates another key, for instance.
02:29:51 Quit evilnick (Quit: Page closed)
02:30:00kugelstill better than doing nothing
02:30:42Blue_DudeI chose view_playlist only because I wanted that key on the e200 target not because I liked the function. I never use it.
02:31:23Blue_DudeI'd rather do something cool like launch the fireworks plugin. :)
02:31:34Blue_DudeNo I'm not going to do that.
02:31:48kugelwell, why not? :p
02:31:53Blue_DudeDoom!
02:32:07Blue_DudeDebug menu, something.
02:32:17Blue_Dudetest_codec.
02:32:19kugelwe could have a hotkey.rock (similar to the autorock.rock which is auto-launched on boot)
02:32:37Blue_DudeI ain't gunna go there, no sirree.
02:35:05 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
02:36:43Blue_DudeOK, so what ought the default be? Still view_playlist?
02:37:08kugelI suppose, since that's the most-stolen button, isn't it?
02:37:35kugelaren't it just 3-4 targets anyway where you steal another buttons?
02:39:33Blue_DudeI haven't stolen *any* other buttons yet, just leaving the possibility to do so.
02:40:02kugelso yes, let it be playlist
02:40:22 Quit merbanan (Ping timeout: 245 seconds)
02:42:59kugelbtw, I find it nice that you (accidentally) finished my initial patch (FS #9873) :)
02:43:10kugelI didn't dare to commit the tree-part of it :p
02:44:04Blue_Dudekugel: I just put this in wps.c around the view_playlist portion. Look right? #if !defined(CONFIG_HOTKEY) || (CONFIG_HOTKEY != PLAYLIST)
02:45:23*kugel wonders how's that supposed to work for other targets
02:45:31kugel*new targets
02:45:45kugelI don't see a point in #define CONFIG_HOTKEY XXX
02:46:27Blue_DudeHm. I was thinking conditional defaults but since we're abandoning that idea...
02:46:39Blue_DudeSave a few bytes? :)
02:46:44Blue_DudeOK, bad idea.
02:53:58 Quit komputes (Ping timeout: 258 seconds)
02:55:39Blue_DudeRefresh my memory, enum don't cost anything if they're not referenced, correct? They don't have to be defined out?
02:57:13Blue_DudeIf a value is left in an enum, or even if an entire enum table is left in, it won't cost anything in bin size?
02:57:32 Join webguest76 [0] (~45448474@giant.haxx.se)
02:58:54Blue_DudeAnyone?
02:59:15 Quit webguest76 (Client Quit)
03:00
03:00:36***Saving seen data "./dancer.seen"
03:00:59kugelenum definitions don't cost anything right, just like struct definitions
03:01:32Blue_Dudeok, thanks
03:01:39kugelbut they do represent a data type which of course costs memory if you alloc an "object" of that type
03:04:27Blue_DudeMy enums only assign values to labels. They don't have names.
03:04:59Blue_DudeSo I think it's OK to take the conditional ifdefs away.
03:09:38 Join hebz0rl_ [0] (~hebz0rl@dslb-088-065-056-058.pools.arcor-ip.net)
03:10:48 Join komputes [0] (~komputes@ubuntu/member/komputes)
03:13:10 Quit hebz0rl (Ping timeout: 252 seconds)
03:17:36 Quit Darkknight512 (Remote host closed the connection)
03:23:18 Quit hebz0rl_ (Quit: Ex-Chat)
03:25:46 Join n1s [0] (~n1s@rockbox/developer/n1s)
03:28:03MinatakuAh, embedded systems. Where people still care about their code (sure, it's because they have to, but still. It's nice.)
03:29:15 Quit Schmogel (Ping timeout: 264 seconds)
03:32:06Blue_DudeOK, latest patch is at FS #11081.
03:32:33Blue_DudeI'd like to commit this one. It's about as clean as it can be at this stage.
03:33:29kugelBlue_Dude: some keymap files have still both?
03:33:45Blue_DudeYeah, the ones that need work. They're not active yet.
03:35:19Blue_DudeOnce the keymaps are fixed, the conditionals can come out.
03:35:58Blue_DudeBTW, that patch has a bug in features.txt. I'm fixing it.
03:39:23Blue_DudeAll right, I'm off for an hour or so. Please take a look. I'll commit later tonight when I have time to handle any issues that come up.
03:40:03 Quit komputes (Ping timeout: 264 seconds)
03:42:16jeffpare there antialiased fonts for rb??
03:43:05n1sno
03:47:58 Join saratoga_lab [0] (~9803c20d@gateway/web/freenode/x-fbghvgjwklsvbywn)
03:49:42 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
03:52:12 Join anewuser [0] (anewuser@unaffiliated/anewuser)
04:00
04:00:49 Join TillW [0] (~Till@h145-net09.simres.netcampus.ca)
04:03:57 Join Barahir_ [0] (~jonathan@gssn-5f75522e.pool.mediaWays.net)
04:07:23 Quit Barahir (Ping timeout: 260 seconds)
04:08:18funmanCGU_PROC is exactly the same in as3525v2 than in as3525
04:14:14funmanso mp3@128kbps decodes at ~30MHz on clip+
04:14:27funmandoes it look sane ?
04:14:37n1syes
04:15:27kugelfunman: I thought test_codec says we're slower?
04:16:11saratoga_labfunman: i'd say not very sane
04:16:17funmankugel: yes because the CPU really runs at 120MHz (240/2)
04:16:29saratoga_labPP takes about 38MHz, and it has basically zero latency memory
04:16:49kugelfunman: did you clock it higher now?
04:17:04 Quit CaptainKewl (Quit: ( www.nnscript.com :: NoNameScript 4.22 :: www.esnation.com ))
04:17:21 Quit arjunvj3 (Remote host closed the connection)
04:17:40saratoga_labthough i guess the ARM9E has a faster multiplier and faster load/store, maybe if the IRAM is also zero latency it could be really efficient
04:18:18funmankugel: trying to
04:18:28funmantest_codec tells me "cannot create logfile" now ?
04:18:36kugelah, the 30MHz are at 120MHz?
04:18:50funmannope, i just divided the 59MHz test_codec previously told me
04:18:58kugelah
04:19:25saratoga_labwith the IRAM running at 65MHz like on the AMSv1, I'd expect something more like 45MHz
04:19:35saratoga_labalso the smurfed out front page is growing on me
04:20:50funmannvm i a new test_codec.rock in apps/ and an obsolete (wrt plugin API) in viewers/
04:21:36saratoga_labfunman: hmm kugel's fuzev1 benchmarks show mp3 being oddly fast on AMS, so maybe it is more reasonable
04:21:48funmansaratoga_lab: do you test different bitrates?
04:22:02saratoga_lab128k and above are all nearly the same
04:22:12saratoga_labits only below that where decoding becomes simplier
04:22:35kugelthey're oddly fast?
04:22:39saratoga_labyeah
04:22:44saratoga_labway faster then I'd expect
04:22:51kugelwhy?
04:23:06saratoga_labvorbis is way faster then mp3 on everything but AMS, where they're basically tied
04:23:54saratoga_lab(ignoring dual core that is)
04:24:19kugelwell, vorbis/wma is rather slow on them
04:25:01saratoga_labon PP vorbis is 25MHz while MP3 is more like 38MHz, then on AMS vorbis and mp3 are both ~38MHz
04:25:42saratoga_labso mp3 stays the same but vorbis get ~50% slower
04:25:46jeffpfunman: i love you! i'm so proud that you and whoever got fuze v2 rb'd!
04:25:47kugelwhy is vorbis so much faster on pp?
04:26:27kugeldoes your mdct get huge improvements simply by fast iram?
04:26:58saratoga_labeven without the new mdct it was already much faster
04:27:14saratoga_labbut i think the new mdct speed up the gigabeat F by a bunch too, and it is arm9 without IRAM
04:27:36kugelbut looking at decode time, samsa is 2x faster again
04:27:53funman29.29MHz for mp3
04:28:17saratoga_labhas anyone ever benched the D2?
04:28:24saratoga_labIIRC its arm9e like amsv2
04:28:37funmansame cpu according to configure
04:29:03kugelsaratoga: can you explain why the MHz is lower for PP even though its decode time is 2x?
04:29:08saratoga_labmp3 does do a lot of single loads and stores because of how its filterbank works out
04:29:20saratoga_labkugel: much lower clock speed
04:29:28saratoga_lab80MHz vs. 250MHz on AMS
04:29:43 Quit planetbeing_ (Remote host closed the connection)
04:30:10funman29.19MHz for 128bkps vorbis
04:30:19kugelbut, samsa clearly decodes the files twice as fast, so the MHz needed for realtime should be lower by about the same regardless of the max clock
04:31:31saratoga_labkugel: the MHz for decode doesn't depend on the max clock
04:31:48saratoga_labif it takes 30MHz for real time it takes 30MHz regardless of the CPU clocking at 3 or 300MHz
04:31:57kugelit doesn't make sense that it needs more (minimum) MHz to decode without stuttering even though it decods twice asfast
04:32:07saratoga_labwhy not?
04:32:11kugelsaratoga_lab: right, that's why it's confusing for me
04:32:58kugel40MHz needed, decode in 30s vs 25MHz, decode in 60s ?
04:33:10saratoga_labno it always decodes at max clock
04:33:17saratoga_labthe test_codec plugin boosts
04:33:51*kugel is maybe too tired
04:34:03 Join Rob2223 [0] (~Miranda@p4FDC9BD2.dip.t-dialin.net)
04:34:09saratoga_labthe MHz column is just clock speed / percentage real time
04:34:19Lloreankugel: 30% of 80mhz is smaller than 15% of 240mhz
04:34:28saratoga_labso it normalizes the overall performance of codec by the cpu clock
04:34:46saratoga_labits not really all that physically meaninfgful actually, but it serves two purposes:
04:34:50LloreanThe percentages are completely made up, but illustrate the point - the time is relative to clock rather than absolute
04:34:58saratoga_lab1) lets us compare decode performance on targets with different clocks
04:36:08saratoga_lab2) provides a number proportional to 1/percentage real time, which is often more intuitive, since percent realtime scales nonlinearly with increasing speed up (e.g the increase from 100% to 200% is much more significant to battery life then from 200 to 400%)
04:37:15 Quit Rob2222 (Ping timeout: 265 seconds)
04:37:26saratoga_labshould I just lock that avi thread so people don't pile on that guy explaining why hes out of his mind
04:38:58 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
04:39:28CIA-5New commit by 03funman (r25413): as3525v2: adjustable CPU freq : CGU_PROC is identical to as3525 after all ...
04:39:48funmanyes, tell him to open a new thread when he's got some code to show?
04:40:09 Quit adnyxo (Read error: Operation timed out)
04:40:41Lloreansaratoga_lab: I'd wait for a last response to my last post
04:41:10LloreanBut if he's still unwilling, lock it with a "come back when you want to be part of the community"
04:41:39 Join webguest32 [0] (~4245faa4@giant.haxx.se)
04:41:41saratoga_labi'd rather just delete it actually and PM him something so that its less confrontational
04:42:03LloreanI don't think my last post was at all confrontational
04:42:30saratoga_labno but even if i lock it other people will still post, since like half our forum can post in locked threads :)
04:42:37kugelfunman: 240MHz? Didn't we run at 248MHz?
04:42:43 Quit JohannesSM64 (Quit: WeeChat 0.3.2-dev)
04:42:59funmankugel: nope we use a fixed setting of 240MHz for PLLA on v2, because we don't know how CGU_PLLA works
04:43:02 Join JohannesSM64 [0] (~johannes@cm-84.215.75.42.getinternet.no)
04:43:17kugelah, the PLL settings are still different, got it
04:43:23Lloreansaratoga_lab: But I think if it's locked before he can respond, that may seem more confrontational to him.
04:43:23funmanit would be easy (although a bit long) to find out, but do we really want to know?
04:43:24saratoga_labsomething is really screwy with AMS performance, I wonder if theres some weird restrictions on IRAM performance or something, like certain accesses are fast and other are slow
04:43:41funman=> modify a bit of CGU_PLLA, measure CPU speed, repeat
04:43:53saratoga_lablike on PP5020 with different IRAM banks having different access times
04:44:51*kugel would really love to re-enable voltage scaling
04:45:22funmansaratoga_lab: we could write a test_memory plugin, but i'm not sure how we could test different banks
04:45:49funmanperhaps define a small block size?
04:45:57 Quit webguest32 (Client Quit)
04:46:07kugelI suspect the delay when waiting for the voltage to be high wasn't enough. we could put a while(delay−−); or poll the voltage register more (so that we don't believe the voltage has stabilized enough until we read it say 5 times)
04:46:27funmankugel: iirc it caused problems with µSD ?
04:47:00kugelyes, but there *must* be some way to make it work
04:48:09saratoga_labfunman: its 5 banks of 64KB, so testing them should be easy, but it seems like a long shot
04:49:04funmanthe difference is the number of cycles required to make a read?
04:49:10kugelif µsd works with high voltage all the time, it must also work with reverting the voltage when accessing the µsd
04:49:43saratoga_labfunman: well its one random theory i just pulled out of my butt :)
04:50:05saratoga_labi should run updated tests on the gigabeat F and see how it compares to AMS
04:50:31funmandidn't amiconn find what happened there?
04:51:17saratoga_laboh on PP5020
04:51:34saratoga_labyeah basically the device has 2 cores, and theres a crossbar switch between them
04:51:49saratoga_labso accessing a core's "local" bank is faster then accessing a remote bank due to some bug
04:51:54saratoga_labthey fixed it in the PP5022
04:52:10funmanoh ok
04:52:36saratoga_labbut something similar could happen on AMS, maybe only one bank can be on the bus at a time and theres a cost to switching them or something like that
04:53:03 Quit kadoban__ (Read error: Connection reset by peer)
04:53:31 Join kadoban__ [0] (~mud@cpe-67-247-80-129.rochester.res.rr.com)
04:54:50 Quit kugel (Remote host closed the connection)
04:59:45 Quit TheSeven (Disconnected by services)
04:59:57 Join The_Seven [0] (~theseven@rockbox/developer/TheSeven)
05:00
05:00:07 Nick The_Seven is now known as TheSeven (~theseven@rockbox/developer/TheSeven)
05:00:40***Saving seen data "./dancer.seen"
05:04:41funmanhm wtf .. my Clip+ shows black screen now..
05:05:38Battousaiwho needs a screen
05:07:53funmansaratoga_lab: the as3525 datasheet says the iram on as3525 is 5 macros of 64kB
05:09:07 Join SirFunk [0] (~Sir@97-92-38-108.dhcp.aldl.mi.charter.com)
05:14:47CIA-5New commit by 03Blue_Dude (r25414): FS #11081 - Hotkey patch. Many targets supported, but some keymaps need work before they can be switched on
05:16:18Blue_DudeBuilds are much faster. More clients now?
05:19:59Blue_DudeAll green. Excellent. Red deltas as expected. It's really easy to see which ones were switched on.
05:22:51CIA-5New commit by 03funman (r25415): as3525*: set up CGU_COUNT register before turning on / modifying PLLs ...
05:23:33funmanpost commit stuck
05:24:56 Quit JohannesSM64 (Quit: WeeChat 0.3.2-dev)
05:29:54saratoga_labroolku seems to have added a half dozen extremely fast machines
05:30:32 Quit anewuser ()
05:31:51Blue_DudeI like 3 minute builds. Very nice.
05:34:40 Quit Blue_Dude (Quit: ChatZilla 0.9.86 [Firefox 3.6.2/20100316074819])
05:34:43funmangrmbl clip+ still doesn't boot, wtf
05:38:49 Quit Horscht (Quit: Verlassend)
05:55:51CIA-5New commit by 03funman (r25416): as3525*: make sure fclk is 24MHz before using it as the clock source for pclk ...
05:58:04funmanit seems i can't reliabily use test_fps to measure pclk, probably SSP is too limiting
06:00
06:06:14pixelma1k for such a change? :/
06:07:36funmanoooh there's no fastbus/synchronous/asynchronous modes in arm926-ejs
06:09:13 Join antil33t [0] (~Mudkips@203-184-54-232.callplus.net.nz)
06:10:08funmanFlynDice_: we need you with a working laptop!
06:10:16 Join joshin [0] (~joshin@174-17-234-237.phnx.qwest.net)
06:10:16 Quit joshin (Changing host)
06:10:16 Join joshin [0] (~joshin@unaffiliated/joshin)
06:13:35FlynDice_funman: I got one today, downloading ubuntu disk as we speak.... I saw your last comment and went looking and I'm quite surprised, I thought I saw the bus mode switching code in the disassembly... You appear to be right though.
06:14:05funmanit doesn't appear to have nasty side effects though
06:14:24funmanbut that also means currently as3525v2 is always boosted
06:15:01funmanwell if i remove those calls => black screen :o
06:16:36FlynDice_actually I kind of hope it prevents writes to the SD........
06:18:24 Quit Hillshum (Ping timeout: 260 seconds)
06:20:00funmanhm if i leave the first change in system_init (bic => fastbus) it boots, and the technical manual says writing as 0 is OK
06:20:33funmanbut cpu is running at 240MHz even if rockbox says unboosted
06:20:45 Join Hillshum [0] (~hillshum@75-165-244-84.slkc.qwest.net)
06:21:43 Join dmb_ [0] (~Dmb@unaffiliated/dmb)
06:23:06 Nick dmb_ is now known as dmb (~Dmb@unaffiliated/dmb)
06:30:50funmanalright i've got it working, just needs to adapt the debug menu o:
06:37:20CIA-5New commit by 03funman (r25417): Fix boosting on as3525v2 ...
06:37:41 Part jeffp
06:37:58funmanFlynDice_: how would you measure pclk ?
06:38:40funmanhm test_fps might just work ..
06:39:53FlynDice_I don't know how to measure it, I just calculated it with plla and the 2 dividers
06:40:02funman870fps at fclk=240MHz, pclk=30MHz; and 108 fps at fclk=30MHz (the 30MHz is supposedly)
06:42:22funman1164 fps at fclk=240MHz, pclk=60MHz; and 290 fps at fclk=60MHz
06:42:41funmanat least there's an effect
06:51:15 Quit joshin (Quit: leaving)
06:53:29 Join JdGordon| [0] (~jonno@110.22.122.97)
06:54:54funmannot much different results between 15 & 30MHz and 40 & 60
06:58:36funmansetting bit 6 of CGU_PERI : data abort
06:59:45funmanwith the as3525(v1) reading it should only make pclk twice lower though
07:00
07:00:42***Saving seen data "./dancer.seen"
07:03:56FlynDice_funman: I could never make CGU_PERI bit 6 1 without a white screen for as3525v1
07:06:36funmanit could be set according to http://forums.rockbox.org/index.php?topic=14064.msg162645#msg162645
07:08:09FlynDice_saratoga said he thought he saw some comments in the as3525 linux patch hinting that this would need to happen in the bootloader.
07:24:15froggymanso how exactly do I smurf my iSmurf?
07:24:20 Quit xiainx (Ping timeout: 246 seconds)
07:28:43funmanfroggyman: you need to download iSmurfSmurfer
07:29:17 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
07:29:47froggymanthats sounds like it might be a virus... what if it makes a mistake while smurfing my iSmurf?
07:39:40saratoga_labfunman: thats the memory clock divider right?
07:39:50funman1<<6 , yes
07:40:19saratoga_labIIRC the linux patch says something like "don't do this while the system is running, it clears dram"
07:40:59funmanoh
07:41:12 Quit xiainx (Ping timeout: 240 seconds)
07:42:25saratoga_labby the way, if AMSv2 can do mp3 decoding at 30MHz, then 25-30MHz is a great choice of the idle clock
07:42:32saratoga_laberr unboosted clock
07:42:43funmantrue
07:43:23funmanany multiple of 240/16
07:44:01funmanthere is someone on ABI forum who measured battery life (14hours with flac), i asked him if he could do it with 60<->240MHz
07:44:33funmanbtw we could support 3 (or more) freqs instead of the current 2
07:45:27saratoga_labyeah thats something we've talked about a lot since a few targets could use it
07:45:51funmannow there are 3 settings but i don't remember the difference between the 2 lowest
07:46:10funmanMAX (easy to understand), NORMAL and DEFAULT
07:46:25saratoga_labIIRC default is only when playback is competely idle
07:46:29funmanDEFAULT & NORMAL are bad names anyway
07:46:36saratoga_labyes
07:46:41saratoga_labdefault rarely happens and so doesn't really matter
07:46:55 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
07:47:18saratoga_laba way to do 25MHz for normal playback, 100 for boosted playback, and 250 for APE and MPEGplayer would be ideal
07:47:25saratoga_labbut i guess thats not too important this early
07:48:02saratoga_labmaybe someday add a codec/plugin api function that toogles between medium and high boost frequencies
07:48:24funmanhm i can measure pclk with the timer
07:55:45 Join Buschel [0] (~ab@p54A3F4C8.dip.t-dialin.net)
08:00
08:01:43 Join phanboy4 [0] (~benji@c-174-49-112-244.hsd1.ga.comcast.net)
08:02:53 Join Zagor [0] (~bjst@rockbox/developer/Zagor)
08:03:15funmanwell it looks alright
08:05:26CIA-5New commit by 03funman (r25418): as3525v2: assume plla is the source for pclk (verified with timer frequency) ...
08:05:41funmanthat means we could tune pclk with test_codec/battery_bench
08:07:38saratoga_labIRAM is driven off pclk like on the AMSv1 players right?
08:08:35FlynDice_I think it's off of hclk which == pclk if it makes a difference
08:11:17saratoga_labthen pclk = core clock probably works best for unboosted, and pclk == high as possible (64MHz?) for boosted
08:12:02 Quit linuxstb (Ping timeout: 276 seconds)
08:12:05 Quit tmzt_ (Read error: Connection reset by peer)
08:12:21 Join tmzt_ [0] (~ircuser@99-157-224-139.lightspeed.bcvloh.sbcglobal.net)
08:14:41 Quit Buschel ()
08:21:07 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
08:23:06 Join ender` [0] (krneki@foo.eternallybored.org)
08:25:41 Quit linuxstb (Ping timeout: 240 seconds)
08:26:10amiconnsaratoga_lab: AMS IRAM isn't single cycle, is it?
08:26:32saratoga_labamiconn: not on the v1 devices anyway, don't know about the new ones
08:26:48saratoga_labbut i guess it can't be boosted since the IRAM clock is only 64MHz IIRC
08:27:22amiconnIiuc it is plain dram which happens to be embedded, and is a bit faster than external dram
08:27:45amiconnTrue single cycle requires sram
08:29:25saratoga_labthe v2 players are weird, they have a ridiculous amount of not very fast IRAM, and then a lot more DRAM
08:31:13 Join evilnick [0] (~evilnick@rockbox/staff/evilnick)
08:31:59funmani should re test *RAM on amsv2, especially with different cpu clk & pclk
08:32:18saratoga_labif iram isn't faster then DRAM maybe its not worth using
08:32:28funmanor put the plugin buffer there
08:32:51funmanor perhaps disable its clock and get more battery
08:33:38 Join mitk [0] (~mitk@195.117.162.130)
08:35:23 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
08:35:56 Join Strife1989 [0] (~michael@adsl-154-2-49.mcn.bellsouth.net)
08:38:57 Quit xiainx (Ping timeout: 260 seconds)
08:39:17 Quit Strife89 (Ping timeout: 276 seconds)
08:40:38 Quit n1s (Ping timeout: 276 seconds)
08:41:24 Join n1s [0] (~n1s@rockbox/developer/n1s)
08:41:30 Join rockabocks [0] (~72f439b2@gateway/web/freenode/x-zclthvrrxipnebrq)
08:41:57rockabockshi, is there an installer available for macs or linux?
08:43:25saratoga_labrockabocks: http://www.rockbox.org/download/
08:43:39 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
08:44:39 Join S_a_i_n_t_ [0] (S_a_i_n_t@203.184.1.208)
08:46:30 Quit rockabocks (Quit: Page closed)
08:46:51 Quit saratoga_lab (Quit: Page closed)
08:47:30 Quit S_a_i_n_t (Ping timeout: 276 seconds)
08:50:31 Quit JdGordon| (Quit: Bye)
08:51:02 Quit xiainx (Ping timeout: 276 seconds)
08:52:12 Join Luca_S [0] (~5d3fc54b@giant.haxx.se)
08:55:25 Join LinusN [0] (~linus@rockbox/developer/LinusN)
08:55:35 Join flydutch [0] (~flydutch@host225-165-dynamic.15-87-r.retail.telecomitalia.it)
08:56:22 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
08:58:37Luca_SFuzeV2: r25418 - got a panic - wait for TRAN state failed (STBY) 1
08:59:52Luca_Splaying from the microSD, playback started fine, pressed the home button, got into the playlist viewer, pressed >>|, the storage icon went on, then panic
09:00
09:00:31Luca_Sthe problem can be reproduced
09:00:43***Saving seen data "./dancer.seen"
09:02:10Luca_SI'd bisect it if there is an archive of previous builds... I don't have a build environment here
09:04:38 Quit n1s (Ping timeout: 248 seconds)
09:07:09 Join petur [0] (~petur@rockbox/developer/petur)
09:10:57funmanthere is http://www.rockbox.org/daily.shtml
09:11:38 Join n1s [0] (~n1s@rockbox/developer/n1s)
09:11:44funmanbtw DeviceChart doesn't have the AMS
09:15:14 Quit xiainx (Ping timeout: 246 seconds)
09:19:24funmanthe guy on ABI will give Clip+ battery bench results tomorrow with adjustable cpufreq enabled
09:20:34 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
09:20:37Luca_Sfunman: the fuzev2 is not on that page
09:21:13funmanhttp://www.rockbox.org/dl.cgi?bin=sansafuzev2
09:23:55funmanhow can we add fuzev2 to the daily.shtml ? i can't read dailymod.pl and i'm not sure the info is there anyway
09:24:42funmanapparently there's a 'usablebuilds' defined?
09:25:13funmanah it's in trunk/ ..
09:26:11Luca_Sr25412 doesn't exhibit the problem, but that is the latest version on the dl.cgi page, thus i can't bisect the exact commit
09:26:33 Quit BHSPitMonkey (Remote host closed the connection)
09:26:49CIA-5New commit by 03funman (r25419): Promote Fuzev2 to unstable
09:27:00 Join efyx [0] (~efyx@lap34-1-82-225-185-146.fbx.proxad.net)
09:27:24 Quit CGL (Quit: Saliendo)
09:28:26mitkkark00fka
09:28:35funmanLuca_S: a quick test, playing from µSD and browsing playlist works on Clip+
09:28:55mitkoops. Sorry, wrong keyboard
09:29:06funmanit could be related to my commits between the revisions you tested, and it could also be a random problem
09:29:17B4gdermitk: time to change password? B)
09:29:48mitkI think so :(
09:29:57Luca_Sfunman: to upgrade, I simply replaced the .rockbox folder. do I need to reload the boot loader too?
09:30:12funmannope, no need to change the bootloader
09:31:18 Quit xiainx (Ping timeout: 245 seconds)
09:32:11Luca_Snow trying to reload the latest revision..
09:33:06funmanalso since the wheel doesn't work you can't change cpufreq setting for example
09:33:33Luca_Sunluckily no
09:33:47funmanhm e200R is in the daily builds
09:34:02Luca_Sisn't it possible to change the lower button to mimic the "scroll down" action, at least temporarily?
09:34:03funmanshould I remove it ?
09:34:47funmanLuca_S: perhaps, but i think it's not important at this stage
09:35:52funmansaratoga: ^
09:36:38Luca_Suhm, r25419: screen corruptions
09:36:57funmanyeah and black screen according to ABI guys :(
09:37:21 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
09:37:40Luca_Sand the panic wait for tran state is still here
09:38:33funmanlooks like i broke pretty much AMSv2
09:40:03Battousaigood thing it's april 1, or people would be upset
09:40:31CIA-5New commit by 03funman (r25420): Remove e200R from daily builds
09:41:02Luca_Stesters are called testers for a reason ;)
09:41:16Battousaii believe they're smurfs today
09:41:37funmanLuca_S: testing the same thing than you on fuzev2: no crashes
09:42:04funmanscreen corruption is there, we probably just need to tweak the delays in lcd driver
09:42:20 Join pamaury [0] (~pamaury@rockbox/developer/pamaury)
09:44:43 Quit xiainx (Ping timeout: 264 seconds)
09:45:09funmanand the clip+ current build doesn't work for me either
09:45:14funmani wonder what's wrong in my test setup
09:45:50funmanusually i just do make bin && cp rockbox.sansa /path/to/.rockbox && umount
09:49:08 Join merbanan [0] (~banan@c-62-220-165-110.cust.bredband2.com)
09:50:28 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
09:50:34funmanbuild from a clean directory just works but the binary is very different from current build (although the difference of rockbox.sansa is only 4 bytes)
09:51:32funmansame arm-elf-gcc/ld , different host gcc
09:51:40Luca_Smaybe some build client has different chain?
09:52:05Luca_Stoo late :D
09:53:20funmansystem_init() is identical for the 2 though
09:54:19 Join stoffel [0] (~quassel@p57B4CF72.dip.t-dialin.net)
09:56:18funmanthe difference is likely due to different offsets of functions being called
09:57:27 Join spaax [0] (opera@85-31-247-126.internetia.net.pl)
10:00
10:01:21 Part spaax
10:09:26funmanhm if CGU_PERI is always based off PLLA and we clear the divider it might be insanely high
10:12:13 Join dfkt [0] (dfkt@unaffiliated/dfkt)
10:17:08 Part froggyman
10:20:45 Quit stoffel (Remote host closed the connection)
10:20:45funmanhu oh µSD error too on Clip+
10:21:23CIA-5New commit by 03funman (r25421): Try to fix problems on Clip+ (not sure why they appear randomly)
10:25:11 Quit Luca_S (Quit: CGI:IRC)
10:28:48 Join JanDo [0] (~JanDo@212.44.150.75)
10:29:31CIA-5New commit by 03funman (r25422): Fuzev2 LCD: replace delays by calls to lcd_delay() (delays not changed)
10:31:16*funman commits without testing
10:31:23 Join Luca_S [0] (~5d3fc54b@giant.haxx.se)
10:31:38 Part JanDo
10:31:40Luca_Snow testing :)
10:31:57funmanyou get also bad colors (pink like) on part of the screen?
10:31:59 Join JanDo [0] (~JanDo@212.44.150.75)
10:32:08 Quit merbanan (Ping timeout: 276 seconds)
10:32:42funmanwhy the green delta ? :/
10:32:53funmanthe function is inline
10:33:16 Quit linuxstb (Ping timeout: 246 seconds)
10:33:51Luca_Sdid it change build client?
10:34:16Luca_Sno more panic it seems, and no screen corruptions
10:34:29Luca_SI stand corrected, screen corruptions are there
10:34:34 Join Jaykay [0] (~chatzilla@p5DC577F6.dip.t-dialin.net)
10:35:03Luca_Sand the panic too
10:35:12funmanno (althought it could have), and the usual random change sare smaller than that
10:35:28funmanscreen corruption = part of screen has a bad color ?
10:35:35funmanconsistent over that part
10:37:02Luca_Sright, it seems pinkish from about two thirds of the screen till bottom, it lasts about a third of a second before returning fine
10:37:10Luca_Sright, it seems pinkish from about two thirds of the screen till bottom, it lasts about a third of a second before returning fine
10:38:58funmanperhaps kugel will have an idea, at least it's not shifted like if we missed some pixels
10:39:11 Join Luca_S_ [0] (~5d3fc54b@giant.haxx.se)
10:40:14 Join Kitr88 [0] (Kitr88@BSN-182-4-208.dial-up.dsl.siol.net)
10:40:59pamauryhuh, someone hacked rockbox ?
10:41:03Luca_S_just a question: why did you say that remapping the "playlist" button to "scroll down" wouldn't be useful now?
10:41:04pamauryrockobx.org ?
10:41:15pamauryThere are 'smurf' everywhere !!!
10:42:22Zagorthose damn smurf hackers
10:42:29 Quit Luca_S (Quit: CGI:IRC (Ping timeout))
10:42:34funmansmurf them
10:42:36 Quit Kitar|st (Ping timeout: 265 seconds)
10:42:46pamaurywho are there ? What does 'smurf' means ?
10:42:54funmanLuca_S_: yes i think it wouldn't be useful now
10:42:59funmanpamaury: smurf = schtroumpf
10:43:09pamauryand how did they do ?
10:43:10Luca_S_pamaury: smurf&um=1&ie=UTF-8&sa=N&tab=wi">http://images.google.it/images?hl=it&q=smurf&um=1&ie=UTF-8&sa=N&tab=wi
10:43:56Luca_S_funman: I could test playing music from the internal storage if I could move the list selection down...
10:44:19Zagorpamaury: a sarsaparilla trojan, I think
10:44:34funmanLuca_S_: true but we can test on other similar devices (Clipv2/Clip+), and i trust kugel for writing scrollwheel support soon
10:44:48funmanZagor: Azrael1.0 ?
10:45:01Luca_S_ahhh, I didn't know he was working on it, sorry ;)
10:45:03pamauryZagor: 'sarsaparilla' ?
10:47:07pamauryOnly the frontpage seems to have changed
10:48:05 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
10:48:34 Quit xiainx (Ping timeout: 268 seconds)
10:50:54 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
10:54:06 Join f_coco [0] (~dcf96389@giant.haxx.se)
10:55:10 Quit phanboy4 (Ping timeout: 265 seconds)
10:56:55 Quit f_coco (Client Quit)
10:56:59 Join dfkt_ [0] (dfkt@unaffiliated/dfkt)
11:00
11:00:44 Quit dfkt (Ping timeout: 276 seconds)
11:00:46***Saving seen data "./dancer.seen"
11:06:15 Quit xiainx (Ping timeout: 252 seconds)
11:11:13 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
11:11:37 Quit funman (Quit: free(random());)
11:12:34 Quit Zarggg (Ping timeout: 240 seconds)
11:13:24 Join Zarggg [0] (~zarggg@65-78-69-194.c3-0.eas-ubr6.atw-eas.pa.cable.rcn.com)
11:14:46 Quit linuxstb (Ping timeout: 248 seconds)
11:15:22 Quit Luca_S_ (Quit: CGI:IRC (EOF))
11:20:11 Join Luca_S [0] (~5d3fc54b@giant.haxx.se)
11:21:20 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
11:26:44 Quit xiainx (Ping timeout: 276 seconds)
11:31:40 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
11:33:06 Join perfectdrug [0] (~marko@p5B0EC16F.dip.t-dialin.net)
11:33:17 Join user124345 [0] (~51e05113@gateway/web/freenode/x-hnvgravewgntvogf)
11:38:40 Quit xiainx (Ping timeout: 240 seconds)
11:41:21 Quit linuxstb (Ping timeout: 264 seconds)
11:41:50 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
11:42:39 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
11:50:14 Quit xiainx (Ping timeout: 268 seconds)
11:50:59 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
11:53:01 Join planetbeing_ [0] (~planetbei@c-71-236-164-204.hsd1.or.comcast.net)
12:00
12:08:20 Join akur [0] (~akur@93.102.82.102.rev.optimus.pt)
12:10:27 Quit Jaykay (Quit: ChatZilla 0.9.86 [Firefox 3.6.2/20100316074819])
12:10:39 Join archivator [0] (~archivato@77.70.28.57)
12:14:44 Part LinusN
12:16:54 Quit archivator (Remote host closed the connection)
12:23:24 Part akur
12:25:28 Join niekie [0] (~niek@CAcert/Assurer/niekie)
12:28:45 Quit xiainx (Ping timeout: 252 seconds)
12:32:52 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
12:35:58 Join krabador [0] (~darkham@host128-183-dynamic.46-79-r.retail.telecomitalia.it)
12:44:56 Quit xiainx (Ping timeout: 264 seconds)
12:47:26pamaurythe website is still smurfing ?
12:48:37 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
12:49:58B4gderit might have something to do with the date
12:51:48pamauryah lol
12:57:56CIA-5New commit by 03funman (r25423): Move fuzev2 to unusable
13:00
13:00:48***Saving seen data "./dancer.seen"
13:03:04 Join archivator [0] (~archivato@77.70.28.57)
13:03:57 Part JanDo (" www.qip.im ")
13:07:13 Join perfectdrug1 [0] (~marko@p5B0EDA4D.dip.t-dialin.net)
13:08:45 Join adnyxo [0] (~aaron@adsl-065-013-002-216.sip.asm.bellsouth.net)
13:09:31 Quit perfectdrug (Ping timeout: 265 seconds)
13:09:36 Quit xiainx (Ping timeout: 245 seconds)
13:10:25 Join JanDo [0] (~JanDo@212.44.150.75)
13:12:59 Join merbanan [0] (~banan@c-62-220-165-110.cust.bredband2.com)
13:14:47 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
13:20:49GodEaterApril Fools Day is not a global phenomenom sadly
13:21:46 Quit blairb (Quit: Leaving)
13:22:02topikfools are though
13:25:16 Quit TheSeven (Quit: ChatZilla 0.9.86 [Firefox 3.6.2/20100316074819])
13:34:14 Nick fxb__ is now known as fxb (~felixbrun@h1252615.stratoserver.net)
13:36:41 Quit xiainx (Ping timeout: 245 seconds)
13:38:16 Join huelk_ [0] (~huelk@dtmd-4db78ea6.pool.mediaWays.net)
13:39:50 Quit huelk_ (Client Quit)
13:39:57 Quit yosafbridge (Quit: Coyote finally caught me)
13:41:59 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
13:54:23 Join Schmogel [0] (~Miranda@p3EE21F35.dip0.t-ipconnect.de)
13:58:33 Nick dfkt_ is now known as dfkt (dfkt@unaffiliated/dfkt)
14:00
14:09:59 Join robin0800 [0] (~quassel@cpc2-brig8-0-0-cust964.brig.cable.ntl.com)
14:10:00 Quit robin0800 (Read error: Connection reset by peer)
14:10:31 Join robin0800 [0] (~robin0800@cpc2-brig8-0-0-cust964.brig.cable.ntl.com)
14:12:35 Quit evilnick (Quit: Leaving)
14:15:04 Join evilnick [0] (~evilnick@rockbox/staff/evilnick)
14:16:25 Quit krabador (Read error: Connection reset by peer)
14:20:10Tornegevaerts: mcuelenaere suggests using the profiling code to implement the boot charting..
14:20:33Tornethis seems rather frightening ;)
14:26:41 Quit xiainx (Ping timeout: 264 seconds)
14:28:26gevaertsTorne: hm, yes... I don't know anything about that code, but it *might* work well
14:29:09TorneWell, I could add -finstrument-functions but not define RB_PROFILE
14:29:18Torneand then implement my own profiling function
14:29:27Tornewhich does rather less than what profile.c does.
14:29:43Tornebut this seems like a disproportionate amount of work, tbh..
14:29:49Torneand it adds a load of overhead at runtime
14:30:10Torneif CHART_CALL is really that objectionable I would probably prefer to just put all the strings in manually, tbh
14:30:17Tornefor the sake of 28 functions it's really not worth it :)
14:30:30Torneit'll only take ten minutes. *g*
14:30:48pamauryTorne: I think I have an alternative solution using C preprocessor trick but I'm trying to make it work
14:30:52TorneOh?
14:30:57Tornepaste what you've got?
14:32:26 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
14:32:35pamauryThe idea is that if you want to log a function "foo" you define: "#define foo(...) (logf("call foo");foo(__VA_ARGS__))
14:32:51pamauryThe problem is that you have to it manually for each function
14:33:08JdGordonI dont tihnk the CHART_CALL() macro is too intrusive
14:33:17JdGordonespecially if we limit to the init funcs
14:33:19Tornepamaury: we considered that kind of thing..
14:33:31TorneJdGordon: yeah, like i said it's used 28 times, only in main.c and settings.c
14:33:38Torneit's not meant to be put *everywhere*
14:33:45Tornebecause if nothing else the logf buffer will wrap
14:34:01Tornei put one on various interesting seemign points in init() and settings_apply()
14:34:05JdGordonI'd still like someone to work on making the apps main() less hardware fidly
14:34:14Torneyes, that would also be good, but unrelated ;)
14:34:20pamauryTorne: and what is the problem with this solution ?
14:34:29Tornepamaury: I wasn't thinking of doing it quite that way
14:34:35TorneI was thinking renaming the symbols, but yeah
14:34:46pamauryThis way you avoid any renaming
14:34:49Tornethe macro will work since macros can't be applied to their own output
14:35:15Torneit also produces nicer logs, actually
14:35:15 Quit Luca_S (Quit: CGI:IRC (EOF))
14:35:34Tornesince CHART_CALL(rc = foo(bar, baz, bash)); actually logs the whole string "rc = foo(bar, baz, bash)"
14:35:47 Join panni_ [0] (hannes@ip-95-222-52-93.unitymediagroup.de)
14:36:57Torneso, hm
14:37:13TorneYou'd end up with a header file contianing a list of htem
14:37:18Torneand you just include that header into main.c and settings.c
14:37:34TorneI guess it might also accidentally decorate some other instances of the same functions in those files? not sure.
14:37:38Tornebut that's probbaly not a huge deal
14:37:46Torneyou can cut the log off once you see it enter root_menu
14:38:07pamaurythe problem with my solution is that it with log all calls, even some you don't want
14:38:19TorneMost of the things being logged are only called once
14:38:22pamauryThe problem with the proposed sqolution is that you need to decorate each call
14:38:31Torneand it won't be all calls, because you only need to include it in a couple of files
14:38:44pamauryYes, all calls within the file
14:38:48Tornebasically everything interesting is called from main(), init(), or settings_apply()
14:38:57Tornei suspect the "collision rate" is virtually zero
14:39:04Tornemaybe even actually zero, not sure.
14:39:09pamauryprobably
14:39:18Torneso yeah, i think it would do
14:39:20Tornei'll try it in a mo
14:40:01TorneYou'd still need some explicit CHART_STR's for a few points that aren't function calls
14:40:05Tornebut htat's ok
14:42:11 Join anewuser [0] (anewuser@unaffiliated/anewuser)
14:42:11pamauryHum, my solution also requires to be careful about definition and declarations, the c preprocessor doesn't make a difference
14:42:57 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
14:43:16TorneYou'd need to include the header last
14:43:33TorneHm, but that might not even be sufficient, some of hte functions init() calls are also defined in main.c
14:43:55pamauryyeah, that's my point
14:44:07TorneIndirect self-references are not expanded, i notice, also, which means you can in fact define foo in terms of CHART_CALL or similar anyway ;)
14:44:08pamauryI don't see a way to get around this problem easily
14:44:14TorneOh, er
14:44:23TorneYour solution also doesn't work for rc = init_foo()
14:44:27pamauryyes it works
14:44:28Tornebecause the macro doesn't expand to an expression
14:44:37pamaury#define foo(...) (printf("You call: " #__VA_ARGS__),foo(__VA_ARGS__))
14:44:47pamaurynotice the parenthesis and the ","
14:44:48 Quit kio (Quit: Leaving)
14:45:04TorneHeh
14:45:08TorneBut then how do you trace afterward? :)
14:45:25Torne(you can do it if you know which functions return values and which are void, i guess)
14:45:42Tornei need a trace before 8and* after, because not everything during init has a trace
14:45:48pamauryerr, you're right, give a minute to try to find another trick
14:45:49Torneso it's not sufficient to measure the time from one thing to the next
14:46:04TorneYou can do it if you have seperate macros for void functoins and int functions
14:46:29TorneOh, no
14:46:35pamauryyes but in one macro :)
14:46:38Torneagain, not easily because you can't put a variable declaration inside an expression :)
14:46:45pamauryyes you can
14:46:46pamaury:
14:46:52Torne...ok, maybe gcc can
14:47:01pamaury({int a; ...; a})
14:47:03Tornebut the idea makes me throw up in my mouth a little
14:47:05Torne:)
14:47:17Tornei guess abusing gcc extensions is fine
14:47:27Tornebut that's still pretty horrifying to look at. but i guess only if you look in bootchart.h
14:47:36Torne:)
14:48:56pamauryTorne: that thing works:
14:48:57pamaury#define foo(...) ({printf("You call: " #__VA_ARGS__);int a=foo(__VA_ARGS__);printf("end of call");a;})
14:49:06TorneYeah
14:49:11pamaurywonderful isn't it ;)
14:49:16JdGordonnot really...
14:49:21 Quit merbanan (Ping timeout: 260 seconds)
14:49:21Tornebut not for funcitons that return anything other than int
14:49:23Tornesurely
14:49:38Tornei'm looking at __builtin_apply and being slightly afraid
14:49:51Tornethat would have to be a functoin, though, which means you are back at symbol renaming ;)
14:51:48pamauryTorne: why ?
14:52:14Torne__builtin_return returns from the containing function with the result of calling the function referred to by __builtin_apply
14:52:27Torneso you'd ahve to rename the real init_whatever to something else and implement a shim init_whatever that used that
14:52:53pamauryno, the macro I have you before works out of the box the function name is foo
14:53:09Tornebut it won't compile if foo returns void
14:53:19Torneso you'd need at least two of them
14:53:27pamauryno, but you just have to have two macros, that's not too much
14:53:35pamauryAnd it will be hidden in a header
14:53:49 Nick fxb is now known as fxb__ (~felixbrun@h1252615.stratoserver.net)
14:54:09Torneright.
14:54:11Tornehm
14:54:13Tornei'll try it
14:54:24Torneit's better than -finstrument-functions ;)
14:57:17 Join merbanan [0] (~banan@c-62-220-165-110.cust.bredband2.com)
14:58:49pamauryTorne: I think there are ugly C extension to do it with one macro, I'm testing
15:00
15:00:49***Saving seen data "./dancer.seen"
15:07:31 Quit linuxstb (Ping timeout: 245 seconds)
15:10:44 Quit xiainx (Ping timeout: 264 seconds)
15:15:45ZagorI just enabled suexec on www.rockbox.org. ping or mail me if something is broken.
15:15:59 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
15:17:18mc2739Zagor: http://www.rockbox.org/download/byhand.cgi is getting Server Error! - Error 500
15:18:08pamauryTorne: I have a solution with one macro only !
15:18:58gevaertsZagor: same for http://www.rockbox.org/mail/
15:20:16 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
15:21:10Tornepamaury: oh? :)
15:21:35pamauryTorne: yes, I can still simplify it but I'll paste it
15:22:16pamaury(5 lines is small enough or I'll get kick by pasting ?)
15:24:47Zagormc2739, gevaerts: thank you. fixed now.
15:24:56pamauryTorne:
15:24:57pamaury#define foo2(...) ({printf("You call: " #__VA_ARGS__);int a= \
15:24:57pamaury __builtin_choose_expr( \
15:24:57DBUGEnqueued KICK pamaury
15:24:57pamaury __builtin_types_compatible_p(__typeof(foo2(__VA_ARGS__)),void), \
15:24:57pamaury (foo(__VA_ARGS__),0),foo2(2,3));printf("end of call");(__typeof(foo2(__VA_ARGS__)))a;})
15:24:58***Alert Mode level 1
15:24:58pamaury:)
15:25:39pamauryoops, s/2,3/__VA_ARGS__
15:27:25 Join leavittx__ [0] (~leavittx@89.221.199.187)
15:27:59 Join User67703 [0] (~theuser12@c39.dnepro.net)
15:29:32JdGordonno..... just no!
15:30:02pamauryoops, other mistake: s/foo/foo2
15:31:07pamauryJdGordon: why ? It's horrible to write but it works nicely, it will be hidden in a header
15:31:18pamauryOtherwise, use two macros
15:32:24 Join einhirn [0] (~Miranda@p5485A3F0.dip0.t-ipconnect.de)
15:33:46CIA-5New commit by 03zagor (r25424): Path change for mod_suexec
15:33:57 Quit xiainx (Ping timeout: 276 seconds)
15:34:09 Quit antil33t (Read error: Connection reset by peer)
15:34:15 Join antil33t [0] (~Mudkips@203-184-54-232.callplus.net.nz)
15:34:28Tornethat's an impressive achievement
15:34:33Tornebut I really dislike it, i'm afraid :)
15:34:58Tornetwo macros is better
15:34:59***Alert Mode OFF
15:35:04Tornei'll have a fiddle, anyway
15:35:11Torne..wait
15:35:15Torneisn't the declarations still a problem
15:35:43pamauryTorne: yes, unfortunately
15:35:58pamauryI can't see a way to get around this problem for now
15:36:03Torneanyway, seriously
15:36:14TorneI think an elegant solution is basically out of the question
15:36:27Tornecpp is just not metaprogrammey enough
15:36:33pamauryyes, two macros seems ok to me
15:36:40Tornetwo macros doesn't fix the declarations
15:37:01Tornei seriously think the choice is between CHART_CALL as-is, and just writing the trace strings manually ;)
15:37:20pamaurythen do it manually if it's just in one function
15:37:24Tornei mean, compliments to you for being able to divine the above macro declaration ;)
15:37:27 Join xiainx [0] (~xiainx@modemcable195.238-202-24.mc.videotron.ca)
15:37:40pamauryno, it's too horrible to be useful, just an amusement :)
15:37:43Tornebut i think actually using it is just too nightmarish, especially if it won't really work because it'll splat the declarations of the functions
15:38:22CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
15:38:22*Torne will just write the traces manually. :)
15:38:36Tornewe've spent far longer than that would take already ;)
15:38:37pamauryIt seems to me that a manual solution is better than the CHAR_ macro which makes it hard to read
15:38:52Torneand while it's more lines, it's obviously ignorable to a reader who doesn't know what it is
15:38:59Tornebecause it obviously looks like trace/debug info
15:40:04Torne(and it makes a slightly better looking chart anywya)
15:41:19 Quit beta2k (Ping timeout: 264 seconds)
15:42:05 Quit JdGordon (Quit: Leaving.)
15:42:24 Join evilnick_B [0] (~0c140464@rockbox/staff/evilnick)
15:47:33 Quit linuxstb (Ping timeout: 265 seconds)
15:53:39 Quit mitk (Quit: Leaving)
16:00
16:04:51 Quit perfectdrug1 (Quit: Leaving.)
16:34:06Tornehm. should bootchart mode automatically enable logf? it's not any use on target othewise, and on the sim it's probably useless to begin with
16:37:42pamauryyou can log it to a file but it makes sense to enable logf
16:38:12Torneit starts before storage_init so it would be a huge pain to log to a file :)
16:38:17*pamaury also notices that logf has an advantage over file logging on target where write is not implemented
16:38:34Torneat th emoment it does _logf or DEBUGF depending if logf is available
16:38:46Tornebut supporting the sim is probably useless.
16:38:52Tornenobody cares about benchmarks there :)
16:39:01Tornei'll just force logf on as well
16:39:51 Join spaax [0] (opera@85-31-247-126.internetia.net.pl)
16:40:16pamauryuse _logf and force to compile with logf enabled, so that you still choose to logf the rest of the file
16:40:37TorneYeah, that's what i was doing
16:40:52pamauryok, perfect
16:44:52Torne..hm, now it doesn't compile
16:45:06*Torne looks to see what he screwed up :)
16:46:53 Join toffe82 [0] (~chatzilla@12.169.218.14)
16:52:21 Quit merbanan (Ping timeout: 276 seconds)
16:52:43 Part spaax
16:58:37 Quit arbingordon (Ping timeout: 246 seconds)
17:00
17:00:52***Saving seen data "./dancer.seen"
17:02:03 Join jgarvey [0] (~jgarvey@cpe-065-190-066-089.nc.res.rr.com)
17:02:07 Join einhirn_ [0] (Miranda@vpn10.rz.tu-clausthal.de)
17:02:12 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
17:02:17 Join JohannesSM64 [0] (~johannes@cm-84.215.116.196.getinternet.no)
17:05:39 Part JanDo (" www.qip.im ")
17:05:42 Join arbingordon [0] (~w@unaffiliated/arbingordon)
17:05:55 Quit einhirn (Ping timeout: 264 seconds)
17:07:31 Join kugel [0] (~kugel@rockbox/developer/kugel)
17:07:39kugel\o/
17:07:43kugelwin! I found the scrollwheel bits
17:09:22User67703
17:09:28User67703oops
17:09:30User67703sorry
17:09:49 Join dmb_ [0] (~Dmb@unaffiliated/dmb)
17:10:17pamaury\o/
17:10:40 Quit dmb (Read error: Operation timed out)
17:12:13 Quit planetbeing_ (Quit: Poof.)
17:18:38*linuxstb wonders if there has been any discussion of http://www.talkingmp3players.com/index.html and if they comply with the GPL.
17:19:34evilnick_Blinuxstb: It was mentioned a while back in -community
17:19:55linuxstbI'll check the logs then... Err...
17:20:03evilnick_BIIRC, AlexP was going to email them and ask them to supply the source
17:21:55kugelit even works :)
17:23:35Torneyeah, their binary downloads just contian the contents of .rockbox
17:23:39Tornenot even the .rockbox folder itself, amusingly
17:23:49Torneand ther'es no mention of source or license anywhere on their site or in the archive
17:24:04Torneso no, they aren't complying
17:24:10 Quit User67703 (Ping timeout: 245 seconds)
17:25:23linuxstbTorne: They make Rockbox itself available for download? I just saw rbutilqt.exe
17:25:37evilnick_Bhttp://www.talkingmp3players.com/support.html
17:25:37 Quit pamaury (Quit: Quitte)
17:25:37linuxstbAh yes, I see them now...
17:25:38Tornethey have "firmware updates"
17:26:54 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
17:26:58evilnick_BAnd also include the most recent bookmark in the "firmware update"
17:27:34kugelwell, it makes me still a bit happy that there are mp3 players sold that ship rockbox no matter of gpl ;)
17:28:03 Quit linuxstb (Quit: Leaving)
17:28:15Torneoh sure, i'm sure they'r enot doing it on purpose
17:28:20 Quit anewuser (Ping timeout: 246 seconds)
17:28:21 Join User67703 [0] (~theuser12@c39.dnepro.net)
17:28:36Tornethey probably haven't modified it even
17:30:10 Quit User67703 (Client Quit)
17:31:07 Join anewuser [0] (anewuser@unaffiliated/anewuser)
17:31:51 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
17:34:33 Join dockimble [0] (~dockimble@77.227.1.24)
17:35:34 Nick Strife1989 is now known as Strife89 (~michael@adsl-154-2-49.mcn.bellsouth.net)
17:35:42linuxstbOn the one hand I'm happy they are offering this service. But we have the issue of out-of-date files being distributed, plus no credit for us etc.
17:36:46linuxstbIt would be nice to open a friendly dialog with them.
17:42:26 Join dmb [0] (~Dmb@unaffiliated/dmb)
17:42:33linuxstbOuch, the "update" is from 21st Jan for all three devices - so not even 3.5
17:43:19 Quit TillW (Quit: This now concludes our broatcast day.)
17:46:53 Quit FlynDice_ (Ping timeout: 246 seconds)
17:47:11 Join FlynDice_ [0] (~FlynDice@c-24-19-225-90.hsd1.wa.comcast.net)
17:51:53archivatordomonoky: did you see fs#11160? :) I was actually surprised how little time it took to rewrite a major part of the processing :)
17:52:15domonokywill take a look.
17:55:45domonokyusing setMaxThreadCount = 1 to make it sequential is a nice idea. :-)
17:56:44*kugel wonders why my changes make the scrollwheel move insanely fast
17:59:07archivatordomonoky: yeah, it simplified things a lot. Goes to say how well-designed Qt really is :)
18:00
18:04:25 Quit bluebroth3r (Ping timeout: 246 seconds)
18:05:11domonokyyou should use the Qt Macros for the flags like its done in the bootloader base class. This way the flags get typesafe.
18:05:31 Quit linuxstb (Ping timeout: 258 seconds)
18:06:19 Join bluebrother [0] (~dom@g224237052.adsl.alicedsl.de)
18:06:19 Quit bluebrother (Changing host)
18:06:19 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
18:06:30archivatorHaven't looked at the bootloader classes, will fix it after the weekend
18:07:59CIA-5New commit by 03kugel (r25425): Fuzev2: Scrollwheel works like a charm :) ...
18:08:22domonokyits just two macro calls and using the newĺy generated type instead of int in capabilitys()
18:08:48 Join mitk [0] (~mitk@chello089078013146.chello.pl)
18:08:57 Join froggyman [0] (~me@unaffiliated/froggyman)
18:09:25domonokyand as you mentioned in the patch tracker, it would be better to move this new pointers into a sub struct.
18:09:47domonoky(in TalkEntry)
18:10:05 Quit petur (Quit: *plop*)
18:11:45archivatorYeah, I'll put in some cosmetic touches like that for the final version. I just wanted to prove to myself that it's doable and, more importantly, improves performance.
18:18:53 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
18:20:13 Join komputes [0] (~komputes@ubuntu/member/komputes)
18:27:25CIA-5New commit by 03torne (r25426): Boot charting support. ...
18:29:22 Part qfr
18:30:33 Join Eugenpaul [0] (~ugnpaul@65.226.32.95.dsl-dynamic.vsi.ru)
18:30:44 Join dockimble1 [0] (~dockimble@77.227.1.24)
18:31:36dockimble1does the fuzev2 scrollwheel work in the latest build?
18:33:54 Quit linuxstb (Ping timeout: 252 seconds)
18:34:49 Quit MagusG (Read error: Connection reset by peer)
18:35:33 Part Eugenpaul
18:35:43 Join Eugenpaul [0] (~ugnpaul@65.226.32.95.dsl-dynamic.vsi.ru)
18:36:00 Part Eugenpaul
18:36:54robin0800dockimble1: this from 5.05 Fuzev2: Scrollwheel works like a charm
18:37:38 Join fyrestorm [0] (~nnscript@static-71-249-251-152.nycmny.east.verizon.net)
18:38:01 Quit dockimble1 (Quit: WeeChat 0.2.6.3)
18:39:33Tornehm, is the build server stuck?
18:39:44Torneit's been doing r25425 for half an hour
18:39:58domonokyTorne: its probably smurfing :-)
18:40:39Torneheh
18:41:58 Join Adubb [0] (~Aldubuc@bas15-toronto63-1279474082.dsl.bell.ca)
18:42:16 Quit archivator (Remote host closed the connection)
18:43:32kugeldockimble: yes
18:44:10dockimbleawesome
18:44:36AlexPevilnick_B: I didn't do anything, but someoone should
18:45:00kugelZagor, B4gder: can you slap the build server a bit, it's lagging
18:46:45 Quit froggyman (Read error: Connection reset by peer)
18:47:27 Join froggyman [0] (~me@unaffiliated/froggyman)
18:48:33 Join dockimble1 [0] (~dockimble@77.227.1.24)
18:50:45 Quit dockimble (Ping timeout: 258 seconds)
18:52:33 Quit antil33t (Read error: Connection reset by peer)
18:52:39 Join antil33t [0] (~Mudkips@203-184-54-232.callplus.net.nz)
18:53:30 Quit slck (Ping timeout: 240 seconds)
18:55:09kugeldockimble1: does the latest build work?
18:55:13 Part froggyman
18:56:37 Join slck [0] (Venci@Slackware.SlackPix.Com)
18:57:08 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
19:00
19:00:55***Saving seen data "./dancer.seen"
19:01:02 Quit Hillshum (Ping timeout: 276 seconds)
19:05:18 Quit dmb (Ping timeout: 246 seconds)
19:06:03 Join phanboy4 [0] (~benji@c-174-49-112-244.hsd1.ga.comcast.net)
19:06:53 Join funman [0] (~fun@rockbox/developer/funman)
19:08:42funmankugel: works if i reduce fclk to 120MHz, else doesn't work
19:09:09 Quit anewuser ()
19:09:31kugelthe delays aren't quite worked out still, it works ok when unboosted
19:10:04funmandid you test with trunk or updated your tree only when you committed?
19:10:24kugelno, trunk worked for me unboosted
19:10:42kugelbut I'm now trying some cleanup and it doesn't work well when unboosted
19:11:01*bluebrother spots a bug entry for something he already worked on and changed
19:11:04funmanalso some reports on ABI that Clip+ buttons don't work
19:11:07kugelhowever when boosted it seems to be fine which is strange
19:11:40kugelwith my cleanup, I get lcd corruption and sluggish UI. when boosted both is fixed
19:11:46*bluebrother grumbles a bit
19:12:04kugelshouldn't it be the other way around?
19:12:23kugeldid we really get the unboosting right?
19:12:46funmanyes
19:13:36kugelI'm wondering why the delays seems to fit when we're running faster
19:14:27funmandelay too long?
19:14:43funmantry highering unboosted frequency
19:14:45kugelit works fine again if I double them
19:15:13funmanhm I suppose we should set the unboosted freq to 24MHz immediately
19:18:42kugelthat will work better?
19:19:10funmanno but it could save some battery (alas measuring it without battery_bench would be tedious)
19:19:28funmanand if we have to tweak the delays in each driver, better do it to the final value now
19:19:50kugelI can hardly tweak if I get something this strange :\
19:19:50CIA-5New commit by 03funman (r25427): fuzev2: enable plugins
19:20:12funmanno problems with the LCD ?
19:20:18kugelheh, I was just about to do that
19:20:22 Join mt [0] (~chatzilla@41.233.151.87)
19:20:47kugelthe UI is very slow when unboosted ..
19:20:57funmanat 60MHz ?
19:21:00kugelyes
19:22:47 Join Tomis2 [0] (~Tomis@70.134.93.53)
19:22:49kugeljudging by test_boost, boosting doesn't really work
19:23:53kugelthe counter isn't any faster
19:24:16CIA-5New commit by 03funman (r25428): Clip+: remove button_hold() : we use software hold
19:25:10kugelah no it seems to be bugged
19:25:18 Quit Tomis (Ping timeout: 268 seconds)
19:25:18 Nick Tomis2 is now known as Tomis (~Tomis@70.134.93.53)
19:26:01funmannow scrollwheel works although i didn't change anything related
19:26:48funmankugel: test_boost works: up to boost, down to unboost
19:26:57funmanbtw it could use rtc
19:27:31kugelfunman: yes but it looks like if you press up twice, and then only once down, it displays normal but still runs boosted
19:28:00funmancpu_boost() keeps an internal counter
19:28:24kugelI know, test_boost is doing it wrong
19:30:29 Quit bmbl (Quit: Bye!)
19:35:04amiconnB4gder/ Zagor: ping
19:37:12B4gderI'm around, but busy...
19:37:21amiconnHanging build
19:37:25B4gderI know
19:39:31 Join pamaury [0] (~c2c7a50a@rockbox/developer/pamaury)
19:40:56 Quit n1s (Quit: Lmnar)
19:44:46 Quit flydutch (Quit: /* empty */)
19:45:20 Join MagusG [0] (magusg@c-76-97-148-35.hsd1.ga.comcast.net)
19:50:37CIA-5New commit by 03kugel (r25429): Fix test_boost boost handling. Also show the number of loops per second.
19:56:13B4gderI suspect Zagor has broken the build server
19:56:49funmanis it really smurfed up?
19:56:57funman(physically)
19:57:03gevaertsThe smurfexec feature?
19:57:27B4gderyes something related to that
19:57:53 Quit phanboy4 (Ping timeout: 265 seconds)
19:58:26funmani wonder if different CPU clocks would make a difference if i leave the Clip+ backlight on
20:00
20:00:32 Join phanboy4 [0] (~benji@c-174-49-112-244.hsd1.ga.comcast.net)
20:02:53 Quit kugel (Remote host closed the connection)
20:03:17 Join Horscht [0] (~Horscht2@xbmc/user/horscht)
20:03:57funmanI start benching Clip+ @60MHz
20:04:20 Part toffe82
20:06:18 Join phanboy_iv [0] (~benji@c-174-49-112-244.hsd1.ga.comcast.net)
20:06:18funmanI just hope it lasts less than 24h, i don't have enough storage for 24h of stereo pcm O:-)
20:06:28 Join Luca_S [0] (~pjIRC@87.18.82.107)
20:06:56Luca_Sthe cgi chat is broken too, not just the build server :/
20:07:27Luca_Stoo bad, I'm eager to try the fuzev2- scrollwheel build :D
20:09:29 Quit phanboy4 (Ping timeout: 265 seconds)
20:20:13 Nick fxb__ is now known as fxb (~felixbrun@h1252615.stratoserver.net)
20:21:37 Quit alexbobp (Quit: leaving)
20:21:44 Join alexbobp [0] (~alex@66.112.249.238)
20:22:28 Join Horschti [0] (~Horscht2@xbmc/user/horscht)
20:24:43 Quit Horscht (Ping timeout: 258 seconds)
20:27:31 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
20:29:17 Join anewuser [0] (~anewuser@unaffiliated/anewuser)
20:29:21 Nick Horschti is now known as Horscht (~Horscht2@xbmc/user/horscht)
20:29:27 Join phanboy4 [0] (~benji@c-174-49-112-244.hsd1.ga.comcast.net)
20:29:55saratoga8GB clips for 35 shipped in the US today on sellout.woot
20:30:10 Quit mitk (Quit: Leaving)
20:32:12 Quit phanboy_iv (Ping timeout: 265 seconds)
20:35:53 Quit kadoban__ (Remote host closed the connection)
20:36:18 Join kadoban__ [0] (~mud@cpe-67-247-80-129.rochester.res.rr.com)
20:38:07 Join dmb__ [0] (~Dmb@198.138.209.17)
20:40:32 Join spaax [0] (opera@85-31-247-126.internetia.net.pl)
20:45:15 Quit Luca_S (Quit: www.pjirc.it by IRCItaly.net)
20:54:25 Part spaax
20:54:41 Nick fxb is now known as fxb__ (~felixbrun@h1252615.stratoserver.net)
20:54:48 Join petur [0] (~peter@d54C6FA7C.access.telenet.be)
20:54:49 Quit petur (Changing host)
20:54:49 Join petur [0] (~peter@rockbox/developer/petur)
20:56:19 Quit petur (Client Quit)
21:00
21:00:56***Saving seen data "./dancer.seen"
21:09:24 Quit phanboy4 (Ping timeout: 265 seconds)
21:11:01 Join phanboy4 [0] (~benji@c-174-49-112-244.hsd1.ga.comcast.net)
21:11:05Strife89Hmmm, my build client lost a log somewhere.
21:11:34Strife89"Server message: fatal build error: missing log file. [This client] has been temporarily disabled."
21:11:41Strife89Any remedies/checks?
21:11:59 Quit phanboy4 (Read error: Connection reset by peer)
21:12:47 Join phanboy4 [0] (~benji@c-174-49-112-244.hsd1.ga.comcast.net)
21:13:03bertrikrestart the client
21:13:43Strife89If that's the best solution, then I'll do it later.
21:13:56Strife89I'd rather not abort these file transfers right now.
21:14:30gevaerts(a) the client, not the PC, and (b) there are problems with the build server right now
21:14:44Strife89Ah.
21:17:33 Quit linuxstb (Ping timeout: 258 seconds)
21:25:00 Join huelk_ [0] (~huelk@2001:4c50:ffff:a9a0:224:21ff:fe8d:132d)
21:27:22 Nick kadoban__ is now known as kadoban (~mud@cpe-67-247-80-129.rochester.res.rr.com)
21:31:16 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
21:32:48 Quit phanboy4 (Read error: Connection reset by peer)
21:36:19 Quit robin0800 (Remote host closed the connection)
21:36:29 Quit linuxstb (Ping timeout: 252 seconds)
21:42:55gevaertsTorne: on my gigabeat F font loading doesn't seem to take very long
21:43:30gevaertsloading sbs and wps however...
21:44:22Torneodd
21:45:05 Quit pamaury (Quit: Page closed)
21:45:33gevaertsah, dircache slows it down
21:46:06Torneyou're using the bootchart?
21:46:41gevaertsgevaerts/logf-dircache.txt">http://www.evonet.be/~gevaerts/logf-dircache.txt and http://www.evonet.be/~gevaerts/logf-no-dircache.txt
21:46:44gevaertsyes
21:47:26Torneholy crap that's slow
21:47:41Tornei guess that's not cabbiev2
21:47:47 Join yosafbridge [0] (~yosafbrid@li14-39.members.linode.com)
21:48:10gevaertsno. It's http://themes.rockbox.org/index.php?themeid=258&target=gigabeatfx
21:48:16gevaertsDo you want results with cabbiev2?
21:48:19Tornebut yeah, wjy does it take 100x longer to load a font for me?
21:48:24Torne:)
21:48:54Tornewell, i don't desperately want any results, tbh ;)
21:49:20Tornei thought it had gotten slower for me too but it turns out i'm mistaken
21:49:37Torneso, i can't really reproduce this slowdown other people are experiencing..
21:49:47Tornefor me it's just dominated by the fact that font_load takes an eternity
21:50:01 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
21:50:05Tornebut it's always done that, i went back nearly 2 years and it made no difference
21:52:16gevaertsgevaerts/logf-cabbiev2-dircache.txt">http://www.evonet.be/~gevaerts/logf-cabbiev2-dircache.txt and http://www.evonet.be/~gevaerts/logf-cabbiev2-no-dircache.txt
21:52:43Tornewait, but cabbie doesn't even have an sbs
21:53:08Tornethe dircache load *really* messes that up
21:53:16Tornewhich it doesn't seem to for me
21:53:34gevaertsyes, three seconds to not load an sbs seems excessive
21:53:35Tornedircache makes no measurable difference to boot time for me, as long as it's not loading dircache in the foreground
21:53:48Tornefor me it's just 6 seconds for font, then 1.5s for wps
21:53:52Torneregardless of dircache
21:54:03gevaertsfragmentation?
21:54:05Torneand then maybe 1.2s for everythign else, total time under 9 secs
21:54:12Tornefor who?
21:54:24Tornemine can't be very fragmented.. i have the first ~500mb of the disk reseved for rockbox
21:54:38Tornei do install/remove builds quite often but i can't see how it could get very fragmented.
21:54:41 Quit xiainx (Ping timeout: 248 seconds)
21:54:54gevaertsMy disk could be fragmented. I actually don't know
21:55:05Tornei'm leaning towards "font_load is provoking worst case behaviour of the sector caching thing"
21:55:16Tornewhich is why i see it on 5.5G and nobody else sees it on other devices
21:55:25gevaertsah, right. That's there as well...
21:55:39Tornebut yeah, maybe your disk is fragmented, i dunno
21:55:41Tornemien isn't
21:56:01Tornei formatted the disk recently, put on a large padding file to occupy the start, rsynced music, then deleted padding and installed rockbox
21:56:10Tornemusic fils ahve never changed, so i should have basically no fragmented files at all
22:00
22:05:32gevaertsNo fragmentation where it could hurt. Some audio files are fragmented, but nothing in .rockbox
22:10:51 Join Tomis2 [0] (~Tomis@70.134.100.179)
22:12:10 Quit huelk_ (Ping timeout: 260 seconds)
22:13:12 Quit Tomis (Ping timeout: 260 seconds)
22:13:12 Nick Tomis2 is now known as Tomis (~Tomis@70.134.100.179)
22:13:33gevaertsWhen was the slowdown rumoured to have occured again?
22:14:42pixelmaaround the addition of sbs
22:15:53gevaertsok, so october more or less
22:17:22gevaertsLooking at my numbers, that could be misleading. When using dircache the sbs load does indeed add 3 seconds, but that's only 0.3 seconds without dircache. The main problem is that dircache slows everything down so much...
22:17:40pixelmamoos named r23258
22:17:45gevaertsOf course I now fully expect someone else to come with another set of numbers where something else is slow :)
22:19:20pixelmaI think Llorean at least has some numbers which pointed towards dircache too but I can't remember if the theory still stands
22:19:56pixelmas/has/had
22:21:10 Join perfectdrug [0] (~marko@p5B0EDA4D.dip.t-dialin.net)
22:22:28 Quit pixelma (Disconnected by services)
22:22:28 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
22:22:44 Quit amiconn (Disconnected by services)
22:22:46 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
22:22:47 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
22:23:08 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
22:23:13Lloreanpixelma: I don't know what the theory is.
22:23:42LloreanI just know that when I disable dirache it boots fast. When dircache does the foreground scan with a splash, it boots fast. When dircache does a background scan during boot, it boots measurably slower.
22:24:21LloreanThis could be because of something else occurring during the scan, or it could be dircaches fault. I don't really have the knowledge about what goes on during boot to hypothesize about causes.
22:24:53LloreanI really don't think the foreground scan should result in a shorter boot though. If that's the case, it should just always do it.
22:26:43gevaertsOK, I have no idea if this is correct or safe, but moving init_dircache(false); to the end of init(); in main.c seems to fix things. It then boots in no-dircache time again, with the disk staying busy for a while longer
22:27:27LloreanSo that just delays the start of the background dircache scan?
22:28:08gevaertsprobably. Well, it does, but I don't know if it has other bad effects
22:29:24TorneLlorean: well, i committed the boot charting patch
22:29:27TorneLlorean: so you can try that? :)
22:30:34LloreanIt sounds like gevaerts may have a solution (or at least, general area to look for one) already. But how do I use it? I can try to get more information later if needed.
22:30:53gevaertsI'm definitely not convinced that this is a solution
22:31:40TorneLlorean: boot charting? it's an advanced option in configure
22:31:49TorneLlorean: hit advanced, then b for bootchart
22:31:51pixelmaLlorean: I wasn't sure if you could still trsce it back to dircache or if you found something else in the meantime or if there was something which mislead you. Maybe I put it wrong but also wanted to get your attention to report yourself :)
22:32:07Tornethen once it boots, it just stores it in logf, you can dump it to a file from the debug menu
22:32:20LloreanSounds plenty straightforward.
22:32:26Tornethat was the idea
22:32:27Torne:)
22:32:44Tornethe last number on each line is the time in centiseconds
22:33:43Lloreangevaerts: Well it seems to "solve" the problem, so it's at least a workaround assuming it's not likely to cause other problems. Who do we ask about that?
22:34:51gevaertsslasheri or pamaury I guess
22:34:58gevaertsSomeone who knows how dircache works
22:35:09Tornewe should probably see if we can trace this back, also
22:35:14Tornewhere by we i mean you
22:35:22Tornesince it doesn't seem to have a problem on my ipod
22:35:28Tornewell not *this* prob blem
22:35:28Torne:)
22:35:43gevaertsDid you post your trace somewhere?
22:35:50LloreanTorne: If it's apparently dircache making other things take longer, that's not new / unexpected, is it? It's just that more other things were added to occur while dircache ran.
22:36:00TorneTrue
22:36:04Tornegevaerts: no, i ahven't posted mine
22:36:10LloreanI'm not sure what you'd trace it back to.
22:36:17TorneLlorean: well, i guess.
22:36:41Torneit makes sense to kick the backgronud stuff off *after* we've done all the foreground stuff..
22:36:49gevaertsWas the WPS loaded at boot back in the pre-sbs days?
22:37:00Tornesince the dircache scan won't've managed to get far enough to actually help by then, prob ably
22:37:03Torneso it's just a hinderance
22:37:22gevaertsWell, it will suddenly start interfering more with buffering if we just postpone it
22:37:37Tornebut it's not postponing it by a lot, though, is it?
22:37:49Torneif loading the sbs/wps only takes <1sec when it's done before dircache..
22:37:50gevaerts10 seconds :)
22:37:55Tornethen we're only delaying it by 1 second
22:38:06LloreanAnd even if it interferes with buffering a little bit, shouldn't music still start more or less immediately? It doesn't wait for a full first buffer, does it?
22:38:07gevaertsyes, true
22:38:24LloreanI'm not sure it matters if the first buffer takes 5 seconds longer, as long as the user doesn't see playback starting noticeably later.
22:38:45LloreanAnd, in this case, possibly earlier because of the shortened boot
22:39:03gevaertsThe basic problem is that we don't have a proper IO scheduler
22:39:34gevaertsAnd no crystal-ball function
22:39:37Torneheh
22:40:28gevaertshm, could the dircache building thread just back off for 100ms (or any other random number) whenever unrelated disk access occurs?
22:40:58gevaertsThat way it should get out of the way immediately during buffering or wps loading
22:45:16*gevaerts isn't sure if he wants to think about the added weird code interdependencies involved
22:45:27 Join Tomis2 [0] (~Tomis@70.134.99.192)
22:46:12 Quit Tomis (Read error: Connection reset by peer)
22:46:12 Nick Tomis2 is now known as Tomis (~Tomis@70.134.99.192)
22:50:02gevaertsBasically store the current tick and current thread from *_write_sectors() and *_read_sectors(), and provide a function to get those, or check them.
22:50:10gevaertsOr possibly do this in the FAT code
22:54:16gevaertshm, we do have storage_last_disk_activity(), so we only miss the thread info I think, unless the dircache code could derive that from looking at the activity numbers
22:54:38 Join Tomis2 [0] (~Tomis@70.134.89.152)
22:55:35 Quit Tomis (Ping timeout: 246 seconds)
22:55:36 Nick Tomis2 is now known as Tomis (~Tomis@70.134.89.152)
23:00
23:00:53 Join piepiepie [0] (~colson@adsl-222-183-164.clt.bellsouth.net)
23:00:57***Saving seen data "./dancer.seen"
23:01:08piepiepieHello.
23:01:22 Quit piepiepie (Client Quit)
23:02:24 Join ClosetotheEdge [0] (~4c402893@gateway/web/freenode/x-ssuuorbjaeocpeyw)
23:02:29 Join xiainx [0] (xiainx@wpa058006.Wireless.McGill.CA)
23:03:02ClosetotheEdgeHey, how's it going? I was wondering, I have a question about downloading a new revision but it seems to be stalled on the website. r25425
23:03:13 Join AlexP_ [0] (~ap@rockbox/staff/AlexP)
23:03:18 Quit AlexP (Read error: Connection reset by peer)
23:03:18 Quit xiainx (Client Quit)
23:03:25ClosetotheEdgeIs there a way I can get the files on the revision otherwise or manually add the changes in this revision to my player?
23:03:55ClosetotheEdgehttp://svn.rockbox.org/viewvc.cgi/trunk/firmware/target/arm/as3525/sansa-fuzev2/button-fuzev2.c?revision=25425&view=markup&pathrev=25425
23:04:02 Quit evilnick_B (Quit: Page closed)
23:05:02 Quit efyx (Quit: Quitte)
23:05:49gevaertsClosetotheEdge: you can build yourself, yes. There's an issue with the build server right now, which should be resolved soonish
23:06:02ClosetotheEdgeHow do I go about building in the changes?
23:07:34 Join kio [0] (~kio@38.98.68.19)
23:07:36gevaertshttp://www.rockbox.org/wiki/SimpleGuideToCompiling
23:07:58kiojust want to say thank you. :)
23:08:00ClosetotheEdgeThank you
23:08:18 Quit dionoea (Ping timeout: 276 seconds)
23:08:20 Join dionoea [0] (~dionoea@yop.chewa.net)
23:10:48 Join mapi [0] (~mapi@KHP222006067242.ppp-bb.dion.ne.jp)
23:12:07 Quit ClosetotheEdge (Quit: Page closed)
23:15:04gevaertswow
23:15:19*gevaerts 's idea seems to work spectacularly well
23:16:35Torneoh?
23:17:49 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
23:18:06gevaertsWell, with the standard build and dircache, you've seen the bootchart. With my earlier change (delay dircache init), it boots faster, but if I then start playback immediately, I still have to wait 10 or more seconds to actually get sound
23:18:26gevaertsWith my new changes, it boots fast, and playback starts right away
23:18:36LloreanThat sounds very hopeful
23:19:01gevaertsHey, where's the "attach file" gone to on flyspray?
23:19:19 Quit anewuser ()
23:19:37gevaertsgevaerts/dircache_backoff.diff">http://www.evonet.be/~gevaerts/dircache_backoff.diff
23:19:42Tornecan we do the same for tagcache load-to-ram?
23:19:50gevaertsOh, definitely
23:20:14gevaertsEarly code without much attention to doing things cleanly, and only for ata, but it works
23:22:41Tornethe time for me to reach playback is substantially more than the time to get to root_menu()
23:22:57Torneso, yeah, i suspect i am also being impeded by this
23:23:05Torneand it just happens that for some reasno the wps load isn't very badly affected
23:23:18 Quit GeekShadow (Ping timeout: 245 seconds)
23:23:28gevaertsWell, have a go with this thing :)
23:23:41Tornelater
23:24:40gevaertsI'd put it on flyspray, but that seems to be a bit broken right now
23:25:07 Join anewuser [0] (anewuser@unaffiliated/anewuser)
23:25:37 Join efgpinto [0] (~dexter@193.136.33.132)
23:26:27*gevaerts uploads a new version that should be fully correct, instead of mostly
23:27:20Tornewell i'm glad i did the chart code now ;)
23:27:20efgpintoHi there... I'm really interested in applying for GSoC at rockbox.. But, I've some doubts about it...
23:27:53gevaertsefgpinto: welcome!
23:28:20gevaertsTorne: well, I don't actually need chart code to see a difference between ten seconds and one :)
23:28:36Tornegevaerts: well no ;)
23:28:57Tornei seriously had no idea it was font loading making mine slow though
23:28:57gevaertsefgpinto: what sort of doubts?
23:29:05Torneand it's at least a hint where to look :)
23:29:06efgpintoIs it possible to test rockbox without having an compatible mp3 or whatever?
23:29:25gevaertsefgpinto: yes, we do have simulators that run on a PC
23:29:47LloreanThey're not emulators though, so depending on what sort of project you're applying for, you may still need a physical player.
23:31:14efgpintooh ok... That nice.. Honestly, I was thinking about one of the secundary projects...
23:32:27efgpintoI feel really confortable with C/C++, so I think those fit better my knowledge and experience..
23:33:35gevaertsWell, most project ideas are C
23:33:51gevaertsWhat did you have in mind?
23:35:29Torneis there a good argment for not turning LBA48 on for all the ATA-based platforms? it takes 300 bytes and it means upgraded disks will Just Work. (also it means you can possibly do larger DMA requests)
23:36:05gevaertsTorne: I think not any more. I'd guess that the argument was mainly that they were pretty uncommon
23:36:21gevaertsFS #11167
23:36:22 Nick dmb__ is now known as dmb (~Dmb@198.138.209.17)
23:36:28 Quit dmb (Changing host)
23:36:28 Join dmb [0] (~Dmb@unaffiliated/dmb)
23:37:36bluebrothergevaerts: what player are the numbers for mentioned in
23:37:46bluebrotherFS #11167?
23:37:50gevaertsbluebrother: ah, right. Gigabeat F
23:38:08Tornegevaerts: i also need to look at if we can detect the broken 5.5G-style hard driv es and not do sector emulation for them
23:38:14Tornegevaerts: but i need to test that
23:38:41bluebrothergevaerts: it should work on any target using ata, i.e. all hdd players, right?
23:38:47gevaertsyes
23:38:59gevaertsWell, except the elio
23:39:03bluebrothernice. Will try to check it out on my hdd targets tomorrow morning
23:39:22bluebrotherespecially the mr100 could be interesting as that has a really slow disk
23:40:46gevaertsIt should be easy to extend to work on all players, but I'm a bit tired right now. I'll try to do that tomorrow evening
23:41:04efgpintogevaerts: I was thinking about... Rockbox Playback Test Driver or Clean up the radio code
23:42:04gevaertsefgpinto: ah, right. I don't know much about those...
23:42:29gevaertsYou need saratoga or JdGordon respectively (unless I remember wrong)
23:47:14gevaertsTorne: if we add this backoff mechanism to more than one thread we have to be careful that they don't start waiting for each other
23:47:16efgpintotks anyway gevaerts
23:47:25Tornegevaerts: Heh, indeed
23:47:38Tornegevaerts: do dircache/tagcache load in parallel, btw?
23:47:44Tornebecause that seems like a bad idea *anyway*
23:47:51gevaertsIf they both load in background, I suspect so
23:47:56*gevaerts doesn't actually know
23:48:00Tornethat doesn't seem like a winning design, if so
23:48:05Torne:)
23:53:19gevaertsMaybe a priority scheme? bool storage_is_busy_elsewhere(int prio) ?
23:53:39gevaertshm, not that easy
23:54:20Tornewell if dircache and tagcache *do* load at once then i would suggest we make them, er, not do that
23:54:23Tornewhich solves the problem :)
23:54:43Tornei would assume that loading dircache is more important, but i could be wrong
23:55:54gevaertsWell, people who use the database can't start playing until the database is ready. I don't know the implications on that from RAM loading
23:56:53gevaertsThe problem I see is that the design basically allows both to suddenly decide to reload things. They currently do that at boot, and presumably on usb disconnect
23:58:05LloreanDoesn't having dircache available make the database scan significantly quicker?
23:58:31gevaertsYes, but it's not only the scan, it's also the loading to RAM

Previous day | Next day