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

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

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

#rockbox log for 2011-09-18

00:01:28 Quit petur (Quit: Leaving)
00:07:00 Join bluefoxx_ [0] (
00:08:11 Quit bluefoxx (Ping timeout: 260 seconds)
00:16:03 Quit bluefoxx_ (Ping timeout: 260 seconds)
00:20:31***Saving seen data "./dancer.seen"
00:22:46 Quit ChickeNES (Quit: Computer has gone to sleep.)
00:26:37 Quit ender` (Quit: I will not engage an enemy single-handedly until all my soldiers are dead. -- Evil Overlord List #115)
00:27:34 Quit stoffel (Remote host closed the connection)
00:28:37 Quit aevin (Quit: «oh, macgyver!»)
00:31:01 Join guymann [0] (~charles@
00:31:36TheSevenTorne: sorry for letting that guy escape from the #freemyipod-support cage :)
00:32:24 Quit factor (Ping timeout: 252 seconds)
00:32:59TheSevenactually iloader is quite a bit faster than the rockbox bootloader on the nano2g, because it replaces an earlier stage of bloaty apple code... this significantly helps in reducing rockbox's boot time, but is neglegible for apple's
00:34:12TheSevenand i'm sometimes even directing some emcore guys towards iloader because it is just more tested and stable compared to emcore (especially due to LCD trouble on some models), even though it isn't being developed any more
00:34:18 Join aevin [0] (eivindsy@unaffiliated/aevin)
00:34:37TheSevenhowever all of those options clearly aren't intended for people who need to ask that kind of questions :)
00:35:27 Quit liar (Ping timeout: 258 seconds)
00:37:06Tornewill we break any players if we start ignoring hte partition types inthe mbr?
00:37:31Torneare there players out there which have valid FAT filesystems on partitions that aren't a FAT type that we don't want to mount?
00:37:57gevaertsI don't think so
00:38:25Tornehm, how do we handle the beast?
00:38:44gevaertsThe first partition is explicitely skipped there
00:39:07TheSeventhe beast will just reformat itself anyway, no matter what you do :P
00:39:17TorneTheSeven: it does not! ;p
00:39:34Torneif you sacrifice a virgin and the moon phases are correct...
00:39:47TheSevenit will, if it's in reformatting mood :)
00:39:56Tornealso I can totally turn the reformatting off
00:40:02Tornei know where the bit for it is
00:40:10 Join natedev [0] (
00:40:11Torneit just means your player becomes a brick instead
00:40:16Tornesince it still doesn't *boot*
00:41:20TorneHm, i guess we should also still ignore type 5
00:41:25 Quit natedev (Client Quit)
00:41:35Torneas that seems like it could have hilarious consequences
00:42:50gevaertsgood point
00:43:30Torneanyway, i'm just gonna change it then
00:43:45Torneif it breaks the world for somebody then, er, y'all know where svn revert is
00:46:00CIA-14New commit by torne (r30566): Allow any partition type, not just known FAT types. ...
00:47:26TorneHm, actually
00:48:06CIA-14r30566 build result: All green
00:48:16tmztwouldn't a white list work better?
00:49:10Tornebecause partition types are irrelevant nonsense
00:49:20Tornethe reasoning is in the full commit message :)
00:49:31CIA-14New commit by torne (r30567): Also remove the fat_partition_types array. ...
00:50:01tmztI don't know every target, but I do know devices exist that save configuration on fat partitions which shouldn't be mounted read only
00:50:17tmztdoes this also mean that rockbox will support more than one partition or card or does it already?
00:50:24gevaertsIt does
00:50:30gevaertsHas done for ages
00:50:37TorneIt's not enabled on all devices, though
00:50:46TorneMost devices just support one, so they stop after monuting the first one found
00:50:47 Quit [Saint_AndChat] (Ping timeout: 258 seconds)
00:51:29CIA-14r30567 build result: All green
00:51:36Tornesupporting more than one requires allocating more ram, and larger code throughout the storage system to take a volume/disk number
00:51:44Torneso it's nicer to not support it on devices that won't need it
00:52:10tmztbut devices with a slot will support multiple partitions on the external card?
00:52:28TorneAnything that supports more than one storage device supports multiple volumes by definition
00:52:32gevaertspossibly. I'm not sure what we set the numbers to
00:53:09Tornefour per drive, unless multivolume is disabled
00:53:17tmztno extended?
00:53:26TorneNo, we don't have any code to parse extended partitions in the first place
00:53:42TorneSo there's no need to support more than four per drive, as we wouldn't be able to find the others anyway
00:54:11Torneso yeah, devices with a cad slot support 8 volumes technically :)
00:54:19Tornefour primary partitions on internal storage, four on a card
00:55:37 Join hilbert [0] (
00:55:54TorneHm, actually it doesn't work that way
00:55:58 Join factor [0] (~factor@
00:56:09Torneit's only four per drive if something explicitly enables multivolume and nothing does
00:56:16Torneso yeah, it's one per drive on all current targets
00:56:26tmztfirst fat partition?
00:56:32tmzthow is this not going to break
00:56:41tmztor it's identified by magic?
00:57:11gevaertsHow is *what* not going to break?
00:57:16tmztI mean if you have partitions without the partition type set will it just try to mount one and stop
00:57:24tmztthis patch to remove the partition type numbers
00:57:27Torneit's identified by successfully mmounting it as a FAT partition
00:59:16TorneFAT doesn't really have actual magic, but it looks at the bpb and does a bunch of sanity checks
01:00:12Torneanyway, unless you know of a player which has a valid FAT partition that's before the data partition, other htan the beast which is already specialcased because it *does* have the FAT partition type, then it's fine, i assure you :)
01:00:14tmztyeah I realised that after I typed that, I was thinking of the DOS4.1 string or whatever on vfat
01:00:23Tornetmzt: yeah that can be anything, though
01:00:27Tornethe system identifier is arbitrary
01:00:36Tornethere are certain common ones but nothing mandatory
01:01:05Tornethe only absolutely fixed thing is the sector has to end in 55 aa but that's also true of MBRs and so makes crappy magic
01:01:48saratogalabkind of confused why someone on the forums wants a way to encrypt his mp3 files
01:02:31Tornesaratogalab: yeah i was gonna point out that this is basically DRM and thus 1) impossible to do in any way that's sane/secure and 2) a NoDo
01:02:36Tornebut then decided i couldn't be bothered
01:04:11piggzso, a firmware update on the daughters ipod gets it working in the new dock as per the manual
01:04:44piggzstrange to use...cant use ipod wheel to control...all controls now through the dock, the ipod just displays a tick
01:06:01 Join [Saint_AndChat] [0] (~Saint]@
01:14:29 Join powell14ski [0] (
01:17:15 Quit bertrik (Ping timeout: 276 seconds)
01:36:22 Quit piggz (Remote host closed the connection)
01:44:21 Join Horscht [0] (
01:44:21 Quit Horscht (Changing host)
01:44:21 Join Horscht [0] (~Horscht@xbmc/user/horscht)
01:44:50 Join Strife89 [0] (~Strife89@
01:45:01 Quit Horschti (Ping timeout: 260 seconds)
01:47:36 Quit Llorean (Read error: Connection reset by peer)
01:50:50 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
01:55:05 Quit pamaury (Remote host closed the connection)
01:55:38 Quit Keripo (Ping timeout: 252 seconds)
01:56:56 Join Horschti [0] (~Horscht@xbmc/user/horscht)
01:57:04 Join Keripo [0] (~Keripo@
01:58:53 Quit Horscht (Ping timeout: 260 seconds)
02:01:15 Quit [Saint_AndChat] (Remote host closed the connection)
02:02:28 Quit Horschti (Ping timeout: 258 seconds)
02:04:01 Join Horscht [0] (~Horscht@xbmc/user/horscht)
02:20:34***Saving seen data "./dancer.seen"
02:24:46 Quit factor (Ping timeout: 260 seconds)
02:30:12 Join factor [0] (~factor@
02:30:25 Quit linuxstb (Read error: Operation timed out)
03:03:30 Quit jae (Quit: leaving)
03:04:03 Join jae [0] (
03:11:50 Quit saratogalab (Quit: Page closed)
03:11:54 Join rbbrittain [0] (
03:14:39 Quit tjb0607 (Remote host closed the connection)
03:16:11rbbrittainTorne: Thanks for removing the MBR check. As discussed in the forums, now we can just tell folks to reformat their SDXC & uSDXC cards to FAT32 without checking if their formatter changed the MBR type.
03:17:10rbbrittainThat means less pressure to add exFAT support (presently a NoDo).
03:17:17JdGordonkugel: ping?
03:17:50 Quit rbbrittain (Quit: CGI:IRC)
03:17:51 Join tjb0607 [0] (
03:25:09 Join ChickeNES [0] (~ChickeNES@
03:44:16 Quit hilbert (Quit: Textual IRC Client:
03:49:35 Join bluefoxx [0] (
04:20:36***Saving seen data "./dancer.seen"
04:24:21 Quit Keripo (Read error: Connection reset by peer)
04:29:56 Join Keripo [0] (~Keripo@
04:35:13JdGordondo we have any hash functions in the core?
04:48:49 Quit amiconn (Disconnected by services)
04:48:50 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:48:50 Quit pixelma (Disconnected by services)
04:48:52 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:48:54 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:49:10 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:54:08 Quit TheSeven (Disconnected by services)
04:54:59 Join [7] [0] (~TheSeven@rockbox/developer/TheSeven)
05:12:29 Join Strife1989 [0] (~Strife89@
05:13:32 Join L-Strife89 [0] (~Strife89@
05:15:48 Quit Strife89 (Ping timeout: 240 seconds)
05:17:44 Quit Strife1989 (Ping timeout: 260 seconds)
05:27:39 Join Rob2223 [0] (
05:32:04 Quit Rob2222 (Ping timeout: 276 seconds)
05:43:53 Join [Saint] [0] (~st.lasciv@
05:48:15 Join [Saint_AndChat] [0] (~Saint]@
05:54:51 Quit [Saint_AndChat] (Quit: AndChat - Android IRC Client -
05:55:20 Join [Saint_AndChat] [0] (~Saint]@
05:55:51 Quit [Saint_AndChat] (Client Quit)
05:56:15 Join [Saint_AndChat] [0] (~Saint]@
06:09:21 Join Keripo1 [0] (
06:09:53 Quit Keripo (Ping timeout: 260 seconds)
06:11:38 Quit bluefoxx (Ping timeout: 260 seconds)
06:12:45 Quit powell14ski (Quit: powell14ski)
06:20:38***Saving seen data "./dancer.seen"
06:39:04 Quit ChickeNES (Quit: Computer has gone to sleep.)
06:58:03 Quit mystica555 (Ping timeout: 258 seconds)
07:25:45[Saint_AndChat]Who manages the commit Twitter feed?
07:26:19[Saint_AndChat]The non-square profile icon looks ugly as sin in the feed.
07:26:36[Saint_AndChat]The clef icon would be more well suited.
07:39:38 Join JdGord [0] (~AndChat@
07:48:48 Quit JdGord (Ping timeout: 260 seconds)
07:50:35 Join JdGord [0] (~AndChat@
07:51:03 Quit L-Strife89 (Quit: To sleep!)
07:58:05 Quit [Saint_AndChat] (Ping timeout: 260 seconds)
07:58:58 Quit [Saint] (Ping timeout: 260 seconds)
08:02:48 Join [Saint_AndChat] [0] (~Saint]@
08:03:12 Nick [Saint_AndChat] is now known as [Saint] (~Saint]@
08:15:13 Quit JdGord (Read error: Connection reset by peer)
08:15:22 Join JdGord [0] (~AndChat@
08:20:39***Saving seen data "./dancer.seen"
08:41:17 Quit JdGord (Ping timeout: 260 seconds)
08:48:46 Join stoffel [0] (
09:16:14 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
09:21:53 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
09:35:01n1sJdGordon: we have a crc32
09:35:53n1sTorne: yes i also noticed the inconsitency of the gnu mirrors when adding some new deps to for the new cf gcc, iirc tha'ts why is hardcoded/default
09:37:07tmztdoes rockbox build with codesourcery for the amsv1 targets?
09:37:15tmztor still only tested with arm-elf
09:37:25n1swe only test our toolchain
09:38:14n1siirc kugle tested with codesourcery's patched gcc at some point and it worked
09:38:29n1skugelp, even
09:38:58n1sit produced slightly slower code than stock gcc though
09:39:32n1syeah, see
09:40:14JdGordonn1s: crc32 is going to be pretty prone to collisions though wont it?
09:40:32n1sor well, the comparison is with gcc 4.0 with the old abi and code sourcery 4.4 with eabi so alittle weird
09:40:39n1sJdGordon: probably
09:40:40JdGordonI want to hash filenames on load
09:41:13JdGordonit might be good enough though, I'll have a looky
09:42:16 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
09:43:15n1sthe core crc32 is also rather slow but for filenames that shouldn't matter
09:44:03n1swhen running a crc check in test codec, the actual crc calculation takes longer than decoding the audio on cf on many targets :)
09:44:30n1smany codecs
09:45:15n1shmm, un-staticing the lookup table would peobably help it for large blocks as it would then be copied to the stack
09:46:46JdGordonI think I have a issue where viewports have their font id's changed without being told, so I need a way to verify that the viewport is using the correct id, I figured hash the filename at load and compare the hash with the fonts hash and do some magic if they dont mathc
09:47:07JdGordonif it slows down font loading a tiny bit it won't be a big deal
09:48:10kugelhow can viewports just change the font is?
09:48:52amiconnn1s: If you want faster crc32 for codec tests, you can use the version with an 8-bit lookup table
09:49:02JdGordonit gets initialised with global_status.font_id[screen] and doesnt reinit when the user loads a different font which gets a different id
09:49:13amiconnCore crc32 is a space saver, using a 4-bit lookup table
09:49:21 Join Stummi|afk [0] (
09:49:37 Nick Stummi|afk is now known as Stummi (
09:49:37 Quit Stummi (Changing host)
09:49:37 Join Stummi [0] (~Stummi@rockbox/developer/Stummi)
09:49:48JdGordonkugel: did you get any further with the buflib resize crash? or will you revert r30480 in the mean time?
09:51:11kugelJdGordon: will have a look later today
09:51:52JdGordonkugel: also, I was oin through the tracker and both your RaaA task and buflib task is still open? forgotten about?
09:59:03n1samiconn: yes, i've been thinking about that but not bothered yet
10:00:33amiconnIirc I did that way way ago for testing the optimized ata code on Players and Recorders
10:01:14amiconnMaybe we should put the 8 bit version into pluginlib
10:01:36amiconnMaybe even put it in core for large memory targets.
10:01:52amiconnThe lookup table is 1KB for the 8 bit version
10:02:06n1swhat does the core use crc32 for?
10:02:27 Join mudd1 [0] (
10:04:39 Join bertrik [0] (
10:04:44amiconnThe settings code uses it to validate the nvram block. The database uses it for something
10:04:53 Quit bertrik (Changing host)
10:04:53 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
10:06:29n1sah, we also seem to have at least 3 implementations in codecs
10:07:31amiconnThe 8 bit version should be more than twice as fast. There are only half as many lookups, and you save the joining/splitting of nibbles
10:08:58n1syeah, the lookups are probably the real killer on cf at least. that database crc doesn't look terribly time critical so i guess we don't need a fast crc in core
10:11:15amiconnThe lookup is probably only a killer on coldfire with its slow dram access and no dcache
10:12:05JdGordonor maybe I should store the *font_id in the viewport instead
10:13:45 Join ender` [0] (
10:19:44 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
10:20:41***Saving seen data "./dancer.seen"
10:35:14 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
10:38:58 Join lovasoa [0] (~lovasoa@2a01:e35:8a2e:8080:e2b9:a5ff:fe5b:ca7b)
10:39:13lovasoaGood morning!
10:41:20lovasoaI just upgraded rockbox to the last svn revision, and some of my skins don't work anymore
10:41:56[Saint]What was the prior revision?
10:42:09[Saint]Check it against the commit logs.
10:42:40JdGordonlovasoa: do your themes load fonts?
10:43:00lovasoaI think so... I'm not sure
10:43:10[Saint]*additional fonts.
10:43:14lovasoaIt's amarock_plus2 for iPod
10:43:24lovasoaand cabbiev2
10:43:44*[Saint] boggles
10:43:46lovasoaamarock_plus2 just doesn't load
10:44:02lovasoaand cabbiev2 make the iPod crash
10:44:17[Saint]Cabbiev2 should. Just use the one for t.
10:44:28JdGordonyour best bet is going back a few days, both issues are being looked at
10:47:03lovasoaSo I don't need to report a bug?
10:48:16kugelJdGordon: is the font_id == handle? or why does it change?
11:01:30 Join ironzorg [0] (
11:05:58 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
11:06:37 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
11:14:03 Join Jerom1 [0] (~jerome@2a02:8420:216:a800:f66d:4ff:fe45:790f)
11:22:20ukleinekwodz: I have another rk27xx player here (from my mother in law, so I must not brick it)
11:22:28n1sthe ffmpeg windowing is faster on the beast already in c with no custom asm at all :)
11:22:55 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
11:22:55ukleinekwodz: rkusbtool -d doesn't work thogh
11:25:13ukleinekRKLDSET.FLW doesn't work either
11:27:13n1si wonder if using ffmpegs smaller buffering scheme is better than the one we currently have
11:27:44n1sit would mean one extra copy but might allow more buffers in iram where that matters
11:29:15ukleinekwodz: the other files listed on the website don't work either
11:29:55ukleinekwodz: it's an Archos Vision
11:34:16 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
11:36:12 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
11:48:35 Quit pamaury (Remote host closed the connection)
11:51:41ironzorganyone with a sansa clip+ here ?
11:52:47ironzorgI've seen a lot of bugs on the forums, and I don't know if its really worth it installing rockbox on it
11:52:59ironzorgI don't wanna brick my only music player
12:01:14JdGordonyou need to be very unlucky ot brick it
12:01:28[Saint]Or very stupid.
12:01:46*[Saint] says this in the nicest possible way.
12:01:54*JdGordon was being polite :)
12:02:17[Saint]You'd have to actively ignore instruction to fuck of up.
12:02:36ironzorgI mean, theres a lot of posts talking about how people have glitches on their sansa
12:02:48ironzorgand it seems that they all followed the instructions
12:03:03ironzorgalso the topmost bug in the bugtracker is a clip+ related one
12:03:21[Saint]Well, sure. But bugs and bricking are totally different things.
12:03:26ironzorgso… I was wondering what my odds are that everything go fine
12:03:48[Saint]If you want PERFECT firmware, don't use anything ;)
12:04:06ironzorgI want a firmware that works at least as good as the one I have now
12:04:30JdGordontry it and see
12:04:31[Saint]Rockbox is in development, but, we offer a release, and a "bleeding edge".
12:04:42[Saint]You probably want the release.
12:04:56JdGordonat the meont you definitly want the release
12:04:57ironzorgyes, my player is supported by the stable version of rockbox
12:05:00JdGordonsvn is a bit dodgey
12:05:18[Saint]You can undo it.
12:05:31[Saint]If you don't like it, uninstall it.
12:05:42ironzorgdoesnt it overwrite the defualt firmware ?
12:06:01*[Saint] suggests reading the manual.
12:07:16ironzorgok thanks, ill rtfm
12:08:41[Saint]There's a dualboot function, so you can easily switch between one or the other to get "the best of both worlds" (if indeed the OF has done value to you)
12:08:54[Saint]...covered in the manual.
12:11:22ironzorgsounds good
12:13:49bertriktechnically, it *does* overwrite the original firmware with both the rockbox bootloader and a copy of the original firmware
12:14:50[Saint]...but doesn't insofar as "its still there".
12:15:23[Saint]It just....molests it slightly ;)
12:16:01JdGordonnew version of FS #12273 just uploaded.... would love some extra testing
12:16:03fs-bluebot use buflib for fonts (patches, new)
12:17:04*[Saint] would love to, but still needs to get a build environment up
12:17:24[Saint]Perhaps a task for this evening.
12:19:00 Quit factor (Ping timeout: 240 seconds)
12:20:45***Saving seen data "./dancer.seen"
12:25:04 Join Buschel [0] (
12:32:20 Quit Horscht (Quit: Verlassend)
12:33:41 Nick evilnick is now known as others (~evilnick@rockbox/staff/evilnick)
12:33:47 Nick others is now known as evilnick (~evilnick@rockbox/staff/evilnick)
12:37:52 Quit [Saint] (Read error: Connection reset by peer)
12:38:29 Join [Saint_AndChat] [0] (~Saint]@
12:38:32 Quit [Saint_AndChat] (Read error: Connection reset by peer)
12:41:30 Quit bug2000 (Remote host closed the connection)
12:45:08BuschelJdGordon: I just tested your latest patch, does not really work either
12:45:11JdGordonBuschel: hey, what theme are you using?
12:45:18 Join webguest37 [0] (
12:45:30Buschelmy own one, shall I upload it somewhere?
12:45:44BuschelI am switching between mine and cabbiev2
12:46:42webguest37do he latest changes to the FAT system affect the usage of CF cards in H1xx series players ?
12:49:41Buschelshould include everything needed
12:49:52JdGordonwhich target?
12:49:53ukleinekgrumpf, now I bricked my mother in laws player :-(
12:49:59Buschelnano 2g simulator
12:50:02webguest37an important issue and no comment ?, scary...
12:50:10 Quit webguest37 (Client Quit)
12:50:10ukleinekit seems the fs is broken
12:50:26bertrikukleinek, sorry to hear
12:51:08JdGordonBuschel: can you just email it please? that link isnt working
12:51:20 Join factor [0] (~factor@
12:51:26ukleineklistening to fm still works, but when I select a file on the flash the player crashes and only reacts on reset
12:52:04ukleinekconnected to usb it doesn't automount anymore
12:52:35ukleinekfsck.vfat takes some time, then tells "open: No medium found"
12:52:40ukleinekany ideas welcome
12:53:19bertrikCan you format it from within the OF?
12:53:53bertrikCan you still access storage as a block device?
12:59:29 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
13:09:36 Join lasser [0] (
13:10:48 Join [Saint_AndChat] [0] (~Saint]@
13:12:34 Join ruskie [0] (ruskie@sourcemage/mage/ruskie)
13:12:34 Quit ruskie (Excess Flood)
13:13:08 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
13:16:00*bertrik is slightly annoyed by code-in-header-file in some parts of the as3525 tree
13:17:22[7]is it at least inlined code?
13:18:33bertrikat least some of it appears to be, yes
13:19:13[7]we really need to switch to -flto or llvm
13:19:34n1slto would be cool
13:19:47 Quit Keripo1 (Read error: Connection reset by peer)
13:20:10bertriksupported by gcc since version 4.5 or so?
13:20:53n1syeah, should work since 4.5, it was experimental in 4.4
13:27:28 Join ruskie [0] (ruskie@sourcemage/mage/ruskie)
13:29:16ukleinekbertrik: there is no option to format in OF
13:29:47ukleinekbertrik: just "reset settings", but that doesn't seem to touch the flash
13:30:19ukleinekbertrik: and yeah, I see /dev/sdb, but when trying to mount or fsck it it goes away
13:30:46ukleinekand after that I need to reset the device
13:31:28pamaurybertrik: which headers have code in it ?.
13:31:35 Join TheLemonMan [0] (
13:31:39 Join fyre^OS [0] (
13:32:03 Join hilbert [0] (
13:32:19[7]ukleinek, bertrik: is that an mmc-based or nand/ftl-based storage system?
13:32:46ukleinek[7]: I'd say nand/ftl-based
13:33:16ukleinek[7]: but just a guess. I have another player that has the same usb vid:pid and that has nand
13:34:22ukleinek$(sudo dd if=/dev/sdb count=1 bs=1 | xxd) already hangs
13:34:24 Quit fyrestorm (Ping timeout: 258 seconds)
13:34:35[7]yeah, and the behavior smells like a bad (or just corrupted?) page not being handled correctly by the ftl
13:34:50[7]i assume you didn't attempt to write to the nand?
13:34:58[7](i mean bypassing the ftl)
13:35:01ukleinekin the of I can see most if not all directories
13:35:27ukleinek[7]: no, just tried utils/rk27utils/rkusbtool
13:35:54[7]so you say you can't even read the first sector via usb, but the device itself can mount the flash just fine?
13:36:03[7]that's smelling more like usb trouble then
13:36:06ukleinekdd no reported "dd: opening `/dev/sdb': No medium found"
13:36:16[7]dmesg output after that?
13:36:56ukleinekand /dev/sdb is gone
13:37:26ukleinekafter unplugging the device it doesn't react anymore to key pressures
13:39:17ukleinekthere is an item "firmware update" in the system settings
13:39:50ukleinekconnecting it via usb after having selected that, reports idVendor=071b, idProduct=3226
13:40:12[7]the dmesg output looks like it's just locking up at some point
13:40:40[7]apparently the partition table read isn't triggering this, but your request is
13:41:16[7]you might want to try accessing it on a non-linux system or removing the ehci_hcd kernel module
13:41:41[7]this looks remotely similar to some trouble i've been having with some ipods
13:42:02*bertrik doesn't understand the rockbox md5sum plugin
13:43:23bertrikthe manual entry is confusing and it doesn't work like the manual describes
13:43:30ukleinekwhen accessing some folders the OF hangs, too
13:52:09 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
13:53:24ukleinek-110 is ETIMEDOUT
13:55:10ukleinek-71 is EPROTO
14:01:41 Join Horscht [0] (~Horscht@xbmc/user/horscht)
14:16:44gevaertshm, do we want a roman numeral tag that converts numbers to roman numerals?
14:18:14*[Saint_AndChat] can't imagine the need arising too often...
14:18:26[Saint_AndChat]A skin tag, one assumes?
14:18:58gevaertsDon't you want your skins to say "Now playing track IX of MMIC"?
14:19:06JdGordonnot really :)
14:19:20[Saint_AndChat]If its not a million LOC, then, I'd see no reason not to.
14:19:37[Saint_AndChat]There's lots of tags that hardly get used ;)
14:19:59bertrikI guess tags are basically only ever added, never removed
14:20:11JdGordontags are incredibly cheap to add
14:20:28gevaertsYes, but they add up. One useless tag is cheap, 50 aren't
14:20:46***Saving seen data "./dancer.seen"
14:20:46gevaertsSo we still need a bit of thinking I'd say
14:20:51JdGordonthere is plenty of bloat elsewhere which people dont care about
14:20:57JdGordoni.e run time screen, debug menus
14:21:00 Quit stoffel (Ping timeout: 240 seconds)
14:21:17[Saint_AndChat]Well, there's hardly fifty, but one day soon ill go through the list and mark off ones to be culled.
14:21:31JdGordonso in principle I have absolutly no issue adding tags which are actually usefl
14:21:39[Saint_AndChat]Things that have more than one way of doing X
14:21:49JdGordonnot gonna hapenn
14:22:09gevaertsThe problem is that it will break skins for no good reason
14:22:31JdGordonstraight replacements can be fixed with a script though
14:22:40JdGordonall the %I* tags are pointless
14:23:05ukleinekhmm, starting sudo ./rkusbtool -s makes lsusb hang
14:23:43[Saint_AndChat]JdGordon: refresh my memory on those?
14:23:50gevaertsIf we ever do a big skin syntax change again, I think we should consider dropping some tags, since all skins would need to be looked at anyway
14:23:52JdGordonnext track *
14:24:09[Saint_AndChat]Why are they pointless?
14:24:33JdGordon%ia(1) == %Ia
14:24:46JdGordonthough it isnt implemented, it would make things cleaner
14:25:12gevaertsJdGordon: I was thinking of the possibility of numbering lists. I'm sure some themes could use that
14:25:14[Saint_AndChat]Not clearer
14:25:49[Saint_AndChat]It certainly wouldn't be clearer, to me at least.
14:26:00JdGordonculling tags is the topic
14:26:11JdGordongevaerts: %Ln i have no problems with
14:26:34JdGordonaltohugh it should be %LN to be inline with %LT and %LI
14:26:50*gevaerts nods
14:27:53[Saint_AndChat]JdGordon: yeah but I was talking about culling tags that already have several ways to do x, not adding a new way to do x with tag y :)
14:28:13[Saint_AndChat]I'm sure there's a few "double ups" in there.
14:28:23[Saint_AndChat]Could be wrong...
14:28:45JdGordonthere might be one or two
14:28:47JdGordonnot many
14:29:07JdGordonanyway, if the topic goes to culling stuff there is far more bloat elsewhere
14:29:10*[Saint_AndChat] fights the urge to add "if any".
14:29:31[Saint_AndChat]JdGordon: oh certainly.
14:30:12 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
14:30:23 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
14:30:26gevaertsThe problem with bloat is that in nearly every case one person's bloat is another person's essential feature
14:31:05*[Saint_AndChat] nods.
14:31:23[Saint_AndChat]The whole skin engine is bloat of you ask some.
14:31:45[Saint_AndChat]Where almost sole reason for lurking this.project.
14:35:34CIA-14New commit by gevaerts (r30568): Add "%LN" tag to retrieve the list item number of the current item. This allows e.g. putting item numbers in skinned lists.
14:36:24JdGordongevaerts: %LN gets the selected items number outside of the lists?
14:36:34JdGordonerr, after the list is drawn
14:37:40gevaertsI only coded that patch, I don't know what it does!
14:38:29CIA-14r30568 build result: All green
14:39:01JdGordonyeah, looks fine
14:40:08JdGordonnow, have fun with a dozen nested %if()'s to do the alternating line drawing thing :)
14:40:14 Quit Jerom1 (Quit: Leaving.)
14:40:33*ukleinek is out of ideas
14:40:35gevaertsI'm pretty sure we need one more patch to do the alternating line thing properly :)
14:40:59JdGordonnope, with %if and %ss it can be done easy
14:41:17JdGordonit will look horrible though :)
14:41:22gevaerts*properly* :)
14:42:20gevaertsOh, I'm not sure your "two items per rectangle" idea is ideal. You need to check if the item exists then
14:42:30 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
14:42:55 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
14:44:31JdGordon%?if(%LN, <, 10)<%?if(%LN, =, 0)<even|%?if(%LN, =, 2)<even|%?if(%LN, =, 4)<even|%?if(%LN, =, 6)<even|%?if(%LN, =, 8)<even|odd>>>>> ?%if(%ss(%Ln, 1,1), =, 2)<even|?%if(%ss(%Ln, 1,1), =, 4)<even|?%if(%ss(%Ln, 1,1), =, 6)<even|?%if(%ss(%Ln, 1,1), =, 8)<even|odd>>>>>>
14:44:56JdGordonthat will only work for the first 99 items :)
14:45:18 Join xDownSetx [0] (
14:45:21JdGordon%LT(1) will give "" if it doesnt exist
14:46:03gevaertsOh, sure, you *can* check. It's just extra ecode
14:46:31JdGordonyou'd do it in two different sub-viewports
14:46:43 Quit Stummi (Quit: Bye!)
14:46:45JdGordonumm, no, that ownt work
14:47:51JdGordonnothing shold crash if the item doesnt exist so it shold be safe
14:48:23JdGordoncompletly changing the topic.... what sort of backlash would there be if I commited the menu shortcuts patch with the main menu item?
14:51:55CIA-14New commit by gevaerts (r30569): Don't use NUM_VOLUMES as array size in sd card drivers. NUM_DRIVES is also wrong (the total number of drives in the system is not really related to ...
14:54:11CIA-14r30569 build result: All green
14:56:41gevaertsAdding a ramdisk will break AMS devices...
14:56:42JdGordonif the last item in a struct is char buf[]; then I can just allocate sizeof(struct)+size and use that buf as a normal array?
14:58:17gevaertsI'm not sure. People used to use that sort of trick all the time in the old days, but I seem to vaguely remember that it isn't correct
15:00:00 Join y4n [0] (y4n@unaffiliated/y4ndexx)
15:00:46JdGordonit compiles.... :)
15:01:19JdGordonlooks like ti works
15:01:36*gevaerts wants to use buflib to get rid of the fixed NUM_VOLUMES
15:01:54gevaertsI'm not sure it's a good idea though
15:02:23JdGordonnow, interesting question. what do I do if two places want to load the same font with different buf sizes? use the bigger one is the obvious choice, but that means unloading the smaller one first perhaps and rejigging handles
15:03:33gevaertsJdGordon: doesn't the trivial solution of just loading it twice *also* mean possibly moving handles (if buflib asks other stuff to shrink and things like that)?
15:04:53JdGordonyeah, but that isnt such an issue
15:05:05JdGordonloading things like fonts is always going to need things to be moved
15:05:14JdGordonI tihnk I'll leave this bufsize stuff for later
15:05:43*gevaerts looks for mini2440 people
15:05:59*JdGordon has one
15:06:02JdGordoncollecting dust :p
15:06:10gevaertsDo you know the sd driver?
15:06:26JdGordonno :)
15:06:30JdGordonI know nuting!
15:07:11gevaertsIt's buggy!
15:08:26gevaertsAnd so is the AMS sd driver
15:10:16gevaertsmaybe not
15:10:27gevaertswell, the mini2440 one is
15:12:04*gevaerts retracts his retraction
15:13:36CIA-14New commit by gevaerts (r30570): NUM_CARDS can definitely be smaller than NUM_DRIVES, so don't pretend that this is an error
15:14:15 Quit Buschel (Ping timeout: 252 seconds)
15:16:03CIA-14r30570 build result: All green
15:16:25 Quit antil33t (Read error: Connection reset by peer)
15:16:43 Join antil33t [0] (
15:18:49 Join MethoS- [0] (~clemens@
15:19:30gevaertsEven sd-tcc780x.c has this issue... NUM_DRIVES isn't necessarily the number of SD cards in the system, it's the number of drives in total. On D2 those *are* different...
15:19:46*gevaerts suspects that we don't have an SD driver without this bug
15:20:46bertrikgo gevaerts !
15:22:06xDownSetxforgive my noobness buys but i saw the pmu init patch for the clip+ and was wondering how i would go about applying it
15:23:23bertrikI think the whole hierarchy of defines surrounding storage is getting a bit complex, at least I don't fully understand it :P
15:24:59gevaertsxDownSetx: may be a good start
15:26:48gevaertsbertrik: the main problem is that lots of storage drivers assume that they are the only storage driver in the system, so they take NUM_DRIVES to be the number of drives they need to handle
15:26:51xDownSetxgreat that's just was i was looking for, thanks gevaerts
15:28:25gevaertsIn many cases that happens to be true (except in some experimental builds with ramdisks and/or loopback devices), but for sd-tcc780x.c it's *never* true, and it still makes that assumption
15:30:58*JdGordon drools over the "export .iso as a usb-cdrom" feature request
15:31:07JdGordonjust randomly throwing that into the mix
15:32:33[Saint_AndChat]Didn't gevaerts half do that USB CD image thing?
15:35:46gevaerts[Saint_AndChat]: I did a quick and dirty proof of concept loopback device patch. That may look similar to the CD thing, except that (a) we probably don't want to do the CD thing as a loopback disk inside rockbox (i.e. we want to isolate it to usb storage), and (b) it didn't do any of the MMC commands you need for a CD
15:35:52[Saint_AndChat]Hit some wall...iirc.
15:39:14*JdGordon wonders why the heck his test isnt loading :/
15:40:46*[Saint_AndChat] needs to play with skinned lists more.
15:40:59[Saint_AndChat]Or, have the time to.
15:41:50 Join bluefoxx [0] (
15:42:12bertrikrecordings from the mic on my sansa clip+ sound horrible, like the audio is modulated
15:43:34*[Saint_AndChat] (just reminded)wonders about FM and recording on RaaA.
15:43:56[Saint_AndChat]PR, of it'll ever happen, rather.
15:45:42pixelmaukleinek: is that thing also the Archos *1* vision? Just curious as I saw one last week, unfortunately I don't think I'll be able to help you with your problem
15:50:32bertrikin the spectrum of a recording on the clip+ there is a wide spike every 1024 samples, also there's a lot of aliasing
15:50:57bertrika bit odd that recording produces a 44 kHz file because the hardware supports only 24 kHz sampling rate max
15:51:47bertrikthe spectrum spike looks like a discontinuity in the waveform
16:00:15[Saint_AndChat]How is it possible for the recording to be 44kHz if the Max sample rate is 24kHz?
16:06:02bertrikmaybe we interpolate during recording, I don't know, might explain the aliasing perhaps
16:09:51 Join freddyb [0] (
16:10:59 Join liar [0] (
16:12:01n1syay, the ffmpeg windowing code is faster on my h300 too, probably since it allows all processing to fit in iram
16:20:49***Saving seen data "./dancer.seen"
16:21:05 Join PugVader [0] (~pug@
16:21:28 Part PugVader
16:28:41 Quit lasser (Ping timeout: 252 seconds)
16:31:27 Join Keripo [0] (
16:31:37 Quit GeekShad1w (Ping timeout: 252 seconds)
16:32:46 Join GeekShadow [0] (
16:38:47 Quit GeekShadow (Ping timeout: 258 seconds)
16:40:38 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
16:52:38*bertrik wonders if he should invent names for PMU registers if the datasheet doesn't mention any, or just use the numeric values
16:53:06bertrikor perhaps just not worry too much and just commit stuf
16:56:00pamauryif you don't really have a clue about the names, just use numeric values
16:56:08[Saint_AndChat]I'd vote for using the values as the datasheet does.
16:56:42pamaury[Saint_AndChat]: the point is that the datasheet doesn't mention them :)
16:56:42[Saint_AndChat]Even if they're made up values, its consistent with *something*.
16:57:17[Saint_AndChat]Oh...I read it incorrectly.
16:57:53[Saint_AndChat]I read it as the datasheet used numeric values when no specific name was given.
16:57:54pamaurybertrik: perhaps numeric values and a comment like "might be related to voltages" or similar
16:57:59[Saint_AndChat]I get it now.
16:58:32 Quit pamaury (Remote host closed the connection)
16:58:33bertrikyeah, I think I'll just mention it in the comment.
17:00:53 Join pamaury [0] (
17:00:53 Quit pamaury (Changing host)
17:00:53 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
17:07:18 Join GeekShadow [0] (
17:18:22 Join zewz [0] (~zewz@
17:18:35zewzHey all!
17:19:29zewzI can't download anything from the server, can anyone fix it?
17:21:11zewz...Or is it just me proving to be just another noob? ;P
17:21:48ironzorgwhats the url you want to download from ?
17:23:31 Join robin0800 [0] (
17:23:39 Quit robin0800 (Read error: Connection reset by peer)
17:23:48 Join robin0800 [0] (
17:24:57 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
17:25:38zewzI first tried the utility at:
17:25:59zewzAnd then I tried the manual way, but both failed
17:26:35zewzAnd all download sites I found were just links to rockbox's site
17:26:44ironzorgyep, it's working.
17:27:58zewzBut since I'm already here at the channel - do you know if it's working on the iriver E10? It's not listed, but I've read that it can work anyway and I really hate the original firmware...
17:29:50LloreanWe list everything we know works, or know is in progress to a certain point. If it's not listed, we don't know, and you're best off going and talking to whoever wrote that.
17:31:32bluebrotherif anyone tells it "can work anyway" that person is lying. At least regarding the official Rockbox version
17:32:16zewzI even found another statement on, and they write that it's just downloading the utility and do it the "normal" way.
17:33:05bluebrotherwow, this is total nonsense
17:33:21zewzHaha, then I got my hopes up for nothing
17:33:21CIA-14New commit by bertrik (r30571): Sansa AMSv2: initialise PMU (power management unit), inspired by how the OF does it ...
17:33:23Lloreanfwpatcher is only for H100/H300 players, right?
17:33:36bluebrotherthat Iriver install guide they link is for the h100, and h100 and h300 are the only ones that did require fwpatcher in the past
17:33:45bluebrotherthese days patching is included in Rockbox Utility
17:34:00bluebrotherso there's no point in using fwpatcher anymore.
17:34:13zewzYeah, their only reference is to a broken link of the manual for h100 anyway
17:34:32bluebrotheralso, they only describe installing the bootloader (if it would work on that player), and installing Rockbox does include installing the real Rockbox on the device as well
17:34:35zewzThe original firmware really sucks
17:34:49bluebrotherthat link works fine for me
17:35:19zewzweird, cuz I still can't download it either
17:35:44CIA-14r30571 build result: All green
17:36:00zewzget this:
17:36:01zewz/tmp/EeMov+RX.pdf.part could not be saved, because the source file could not be read.
17:36:01zewzTry again later, or contact the server administrator.
17:36:20zewzAnyway, thanks for the comments
17:36:32zewzMy neighbor just cooked some food for me =)
17:37:29bluebrotherwow, it gets even worse:
17:38:22bluebrotherand to uninstall you should start Rockbox Utility then press "u" to uninstall? Now even mixing up sansapatcher with Rockbox Utility :(
17:39:09bluebrotherHmpf, even after allowing all those JavaScript I can't comment on that.
17:43:18 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
17:43:46 Join mamarley [0] (~quassel@2001:470:5:84d:221:6aff:fe19:4d0c)
17:46:56 Quit mamarley (Remote host closed the connection)
17:52:08 Join lasser [0] (
17:52:48 Join Buschel [0] (
17:53:19 Quit lovasoa (Remote host closed the connection)
17:59:07[Saint_AndChat]That last kink is hillarious.
17:59:45*[Saint_AndChat] gives up and slinks away to a corner.
18:03:39[Saint_AndChat]It has me wondering how many other hilariously broken and just plain wrong Rockbox install guides exist out there in the internet wilds...
18:05:32 Join Jerom1 [0] (~jerome@2a02:8420:216:a800:f66d:4ff:fe45:790f)
18:05:58Llorean[Saint_AndChat]: It's probably best not to ask. Just youtube alone probably has more than you can go through in a sane amount of time.
18:08:11[Saint_AndChat]Thankfully the PSgroove hype seems to have died down. I recall commenting on many a wildly inaccurate "guide" during such times.
18:08:46[Saint_AndChat]And lots of irate users finding no support and eventually ending up here.
18:09:14 Quit robin0800 (Quit: Leaving)
18:09:18 Quit freddyb (Quit: Leaving)
18:16:42 Quit Buschel (Ping timeout: 260 seconds)
18:16:51CIA-14New commit by bertrik (r30572): Add the plugin_start prototype to plugin.h
18:19:19CIA-14r30572 build result: All green
18:20:52***Saving seen data "./dancer.seen"
18:31:00 Quit lasser (Ping timeout: 240 seconds)
18:32:21 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
18:35:36 Join powell14ski_ [0] (
18:36:30 Quit zewz (Quit: Leaving)
18:38:21 Quit powell14ski_ (Client Quit)
18:46:07 Join Stummi [0] (~Stummi@rockbox/developer/Stummi)
18:46:31 Join Keripo1 [0] (
18:47:07 Quit Keripo (Ping timeout: 260 seconds)
18:54:05 Join lorenzo92 [0] (
18:55:33 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
19:09:44 Join Casainho [0] (~chatzilla@
19:16:15 Quit pamaury (Read error: Connection reset by peer)
19:16:34 Quit lorenzo92 (Quit: ChatZilla 0.9.87 [Firefox 6.0.2/20110905175240])
19:45:34 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
19:48:26 Quit Zarggg_ (Remote host closed the connection)
19:48:53 Join Zarggg_ [0] (
19:49:50 Nick Zarggg_ is now known as Zarggg (
19:55:57 Quit fs-bluebot (Ping timeout: 260 seconds)
19:56:10 Quit bluebrother (Ping timeout: 260 seconds)
19:57:10LalufuTorne: this is what an iPod touch looks like by default on an USB port:
19:57:58Lalufuthis is what it looks like with an 191k ohms resistor on the acc-detect port:
19:58:16Lalufu191k is used to signal that the device would like to use IAP via USB
20:05:02 Join petur [0] (~petur@rockbox/developer/petur)
20:07:42 Join fs-bluebot [0] (
20:08:53 Join bluebrother [0] (
20:08:53 Quit bluebrother (Changing host)
20:08:53 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
20:19:57 Join Buschel [0] (
20:20:54***Saving seen data "./dancer.seen"
20:31:33 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
20:31:55 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
20:34:27 Join froggyman [0] (~seth@
20:34:27 Quit froggyman (Changing host)
20:34:27 Join froggyman [0] (~seth@unaffiliated/froggyman)
20:38:32 Quit Buschel (Ping timeout: 260 seconds)
20:38:55 Quit Stummi (Quit: Bye!)
20:40:43 Join Strife89 [0] (~Strife89@
20:40:57 Quit hilbert (Quit: Textual IRC Client:
20:41:56 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
20:41:56 Join Buschel [0] (
20:41:57 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
20:44:21 Join hilbert [0] (
20:48:57 Join fml [0] (
20:50:08fmlHas it been considered to allow a negative start index in the %ss WPS tag? A negative index would mean "n-th character from the end" (-1 would mean the last char)?
20:51:09fmlgevaerts: is the list number 0 or 1 based?
21:08:50 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
21:08:50 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
21:22:27 Quit fml (Quit: ChatZilla 0.9.87 [Firefox 6.0.2/20110902133214])
21:24:29 Join ChickeNES [0] (~ChickeNES@
21:25:04 Quit otih (Remote host closed the connection)
21:25:42 Quit Buschel (Ping timeout: 260 seconds)
21:37:41 Quit ChickeNES (Quit: Computer has gone to sleep.)
21:43:54 Join knob [0] (~knob@
21:44:28 Quit MikeH__ (Ping timeout: 276 seconds)
21:45:48 Part knob
21:46:16 Quit Jerom1 (Quit: Leaving.)
21:46:41 Quit liar (Read error: Operation timed out)
21:47:06 Join MikeH__ [0] (
21:48:12 Join liar [0] (
21:50:46 Join ChickeNES [0] (~ChickeNES@
21:52:02 Quit [7] (Remote host closed the connection)
21:54:08 Part liar ("hallowed are the ori!")
21:59:02 Quit TheLemonMan (Quit: WeeChat 0.3.5)
22:03:22 Quit benedikt93 (Quit: Bye ;))
22:19:21 Join stripwax [0] (
22:20:57***Saving seen data "./dancer.seen"
22:23:52 Join lasser [0] (
22:26:44 Quit petur (Quit: Leaving)
22:35:21 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
22:46:37 Quit TheSeven (Read error: Connection reset by peer)
22:58:10saratogan1s: on later ARM devices the whole windowing/imdct code is kind of ugly
22:58:47saratogamdct in particular is optimized to fit everything into IRAM but has terrible access patterns so on devices without iram its pretty bad
22:59:54 Quit y4n (Quit: only amiga makes it possible)
23:05:48 Join stoffel [0] (
23:07:45 Quit stoffel (Remote host closed the connection)
23:11:59 Quit lasser (Quit: ChatZilla 0.9.87 [Iceweasel 3.5.16/20110815153231])
23:16:52 Quit xDownSetx (Quit: Leaving)
23:26:36 Quit bertrik (Ping timeout: 260 seconds)
23:29:15 Quit mudd1 (Ping timeout: 252 seconds)
23:33:36amiconnsaratoga: Depends on how large the (data) cache is
23:35:12saratogaIIRC the MDCT look up table is 16 KB alone, so unless you've got 16KB you're probably going to flush it between L and R channels
23:35:39 Quit Strife89 (Quit: Heading out.)
23:36:09saratogaoh no its only 8KB
23:37:40saratogastill since accesses are always non sequential thats going to be slower then it needs to be on the beast
23:41:23 Join mudd1 [0] (
23:43:51 Quit TBCOOL (Quit: maintenance. up 32d2h.)
23:44:19 Quit Casainho (Quit: ChatZilla 0.9.87 [Firefox 6.0.2/20110902133214])
23:49:19n1ssaratoga: yeah i've noticed that, what i'm doing now is transplanting ffmpeg's windowing/overlap add code to tremor
23:49:33saratogastripwax and i argued about that ages ago
23:49:37n1sso not touching the mdct itself
23:49:44saratogai think he convinced me that what we had was better but i forget
23:49:50n1sit works and gives a nice speedup on the beast
23:50:23n1sit's also better for targets with small iram as all the buffers it can use now fit
23:50:23saratogawhats teh advantage?
23:50:39n1smuch less mem access basically
23:50:42stripwaxmm.. hello everyone.. let me scroll up and read some more...
23:50:44saratogaIIRC the problem with it was that it had a weird memory access pattern that prevented using ldm on armv6
23:51:52n1sthey calculate half the mdct into one 1/2 block buffer and overlap add that with windowing on the fly to the last half to create the output
23:52:18stripwaxI think the memory access pattern is pretty linear .. I recall you were arguing that it saved Mults and I was arguing that it couldn't possibly save Mults..
23:52:22n1swhat tremor does now is calculate the same half mdct, then double it, then window it, then overlapå add
23:53:21n1syeah theaccesses are linear
23:53:36n1si'll try to clean it up and post a patch during the week
23:53:44stripwaxI think you only save from the reading+writing and then re-reading again later. you still have to read the input twice though (I think) as the window is nonsymmetric
23:53:54saratogaah ok i'm probably forgetting what we talked about
23:54:00stripwaxso I think 'much less' is maybe 'a bit less'
23:54:01n1sstripwax: yeah
23:55:21stripwaxwe should also (if someone has spare time) try and see if changing the mdct to use a real bitreverse format (instead of the odd digitreverse that it currently has) would improve anything at all. For example, it would enable us to remove entirely the digitreverse index array.
23:55:34n1sif i'm not too tired this saves about 2-3 reads and writes per sample
23:56:02stripwaxah is that just because the way I ended up doing the reflection is a bit crap?
23:56:37stripwaxe.g. a better way of doubling-up the output might give you a reduction too
23:57:23n1snot just because of that, tremor now reads in the doubled mdct to do windowing and the writes it back and then reads again to do the final adds
23:58:19stripwaxah true. so, I'm trying to remember why the official Tremor trunk doesn't do this, but the official "low mem" Tremor branch does..
23:58:53stripwax(also, noticed you said you're porting from ffmpeg, but maybe porting from lowmem tremor might be easier - although maybe not of course -)

Previous day | Next day