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 2010-11-21

00:00:56 Quit Keripo (Read error: Connection reset by peer)
00:00:59saratogathe mdct windowing code in libmad looks a bit weird, so maybe we could save another MHz in there
00:01:16saratogabut after that I have no other ideas how to make libmad faster on arm9 (non-E)
00:02:28kugelsaratoga: redownload
00:03:01 Join liar [0] (
00:04:34kugelmost codecs under 30MHz needed
00:05:44 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
00:05:48 Join Keripo [0] (
00:10:43kugel7% cpu usage while decoding vorbis
00:17:48saratogawow, android devices must have faster memory then mp3 players
00:18:18saratogafor fun, you could try aligning all the members in the structs in the libmad header files (mpa.c, frame.h, etc)
00:18:24saratogai bet that makes mp3 a lot faster still
00:19:26saratogavorbis takes 20 MHz, and we haven't even started working on arm11 optimizations for it
00:34:18 Quit Kupop (Ping timeout: 265 seconds)
00:39:09 Quit factor (Ping timeout: 276 seconds)
00:52:11 Quit {phoenix} (Remote host closed the connection)
00:52:30*TheSeven spots a nice color pattern on his ipod classic's LCD
00:53:24*TheSeven needs more ipod classics to test the different lcd types
00:53:46***Saving seen data "./dancer.seen"
00:59:42TheSevenanyone around who knows the arm9e MMU well?
01:00:13TheSeveni just want it to map everything 1:1, to enable the data cache
01:00:17TheSevenwhat do i need to do?
01:01:10TheSeveni'm currently writing 0x???00c1e to the cacheable and 0x???00c12 to the noncacheable section entries of the top level page table
01:01:28TheSevenhowever something is locking up, and it's rather hard to debug why exactly
01:01:57TheSevenas soon as I set CP15 r1 to 0x0005107D it breaks
01:06:41 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
01:09:30 Quit pamaury (Remote host closed the connection)
01:10:18 Quit dfkt (Quit: -= SysReset 2.53=- Sic gorgiamus allos subjectatos nunc.)
01:10:45 Join Buschel [0] (
01:11:57 Quit Buschel (Client Quit)
01:12:59TheSevenkugel? Torne? who else knows these ARM cores well?
01:13:24Torneer, i would help but i am way too drunk. check back tomorrow ;)
01:13:25saratogafunman, flyndice, kkabujan, etc
01:13:35 Quit BHSPitMonkey (Remote host closed the connection)
01:13:56 Join advcomp2019 [0] (~advcomp20@unaffiliated/advcomp2019)
01:14:08TheSevensaratoga: too bad none of these are here :(
01:14:21saratogayou could ping jhmikes
01:15:03saratogai think he wrote all that for arm11
01:15:05gevaertsOr send strong coffee to Torne
01:15:25saratogaalthough if you're just got a generic arm9e core, you might be able to copy and paste the AMSv2 code
01:15:51TheSevenis it in their crt0.S?
01:17:20TheSevenah, memory_init.S
01:17:31 Quit GeekShadow (Read error: Connection reset by peer)
01:22:36 Quit Llorean (Disconnected by services)
01:22:48 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
01:23:45saratogaIIRC on AMS we just move the DRAM and IRAM together in the address space, disable paging, and then enable the caches
01:23:47soapHow many people just got a PM from phly95?
01:24:04saratogai didn't, guess i'm not special
01:24:55 Quit ender` (Quit: Jesus may be able to walk on water, but Chuck Norris can swim through dry land.)
01:24:56saratogahis only post is some crap i forgot to delete, so hes a zero poster again
01:25:18soapYes, he appears to be unable to read how to post.
01:26:13soapIn the USA, so likely not a language barrier. Oh well.
01:26:48soapSorry for interrupting real discussion.
01:26:56*TheSeven found the problem
01:27:05TheSevendamn alignment constraints
01:27:15 Join Buschel [0] (
01:27:34 Quit Llorean (Read error: Connection reset by peer)
01:32:03Buschelthe aligning seems to make some difference ;)
01:32:10saratogayeah on arm11
01:32:15saratogaprobably not much on arm9
01:32:32saratogaBuschel: if you have any ideas how to speed up libmad on arm9 I'd love to hear them
01:32:38 Join Llorean [0] (
01:33:05 Quit Llorean (Changing host)
01:33:05 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
01:33:58Buschelif I should find some new stuff I'll for sure let you know :)
01:34:17saratogai had another idea for arm9e, so i'm going back to the c code for a bit
01:35:20Buschelfor now I think that the stall-reducing asm you've pasted is the next step that does not need changes in algorithm
01:35:53Buschelbtw, I've updated the alignment patch some minutes ago. if this is agreed here I can submit it tomorrow
01:36:24saratogai'm ok with whatever, just get it in
01:36:46 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
01:37:03saratogaFWIW i'm now convinced that using arm9E we should be able to get the entire synth filter down to about 4MHz with optimal scheduling and load/store magic
01:37:09BuschelI'll wait for any review comments and submit tomorrow, if there are no "nays"
01:37:55Buschelonly the dewindowing? or also including the dct stuff?
01:38:02Buscheleitherway -> its _fast_
01:38:14saratoganot including the DCT
01:38:34Buschelwell, I might adapt mpc to such solution then :)
01:39:16saratogain theory you can manage 1 clk for the mul, 1/2 for the packed D coeficient, and 1/2 for the sample data (assuming you can keep it in register long enough to compute two output samples with it)
01:39:31saratogaso thats about 2 clocks per filter tap, plus a bit of overhead for other things
01:41:07saratogaon arm 11 its even better, since you can do most of your loads a single cycle 64 bit double word load, although ensuring alignment for those will be difficult at best
01:41:09Buschelbtw, you've talked about the codec efficiency on the nano 2g some hours ago. the 48 MHz clock is more efficient (mp3 = 43 MHz, mpc = 26 MHz) as the 192 MHz clock (mp3 = 52 MHz, mpc = 33 MHz).
01:41:35saratogai think both tests were done boosted, so 192 mhz clock right?
01:42:01saratogaanyway, shit i'm late to dinner, have to run
01:42:21saratogawill check the logs tomorrow
01:42:36Buscheli will also get some sleep now! see you tomorrow
01:42:51 Quit Buschel (Quit: ChatZilla 0.9.86 [Firefox 3.6.12/20101026210630])
01:50:18 Quit bertrik (Quit: :tiuQ)
01:55:46 Part domonoky1
02:02:55 Join InsDel [0] (
02:03:06 Quit DerPapst (Quit: Leaving.)
02:37:22 Quit casainho (Quit: ChatZilla 0.9.86 [Firefox 3.6.10/20100922073514])
02:53:49***Saving seen data "./dancer.seen"
03:10:23TheSeventhose ipods are insane
03:10:36TheSeven16 DMA channels, powered by a total of 4 AHB masters
03:29:06TheSevenit even has memory to memory DMA
03:32:26kugelmem2mem dma is special?
03:34:46kugelthe old pl081 found in as3525 (and v2) has it too
03:35:19 Quit kugel (Remote host closed the connection)
03:52:28 Quit Gnea (Ping timeout: 240 seconds)
04:03:14 Quit bluebroth3r (Read error: Connection reset by peer)
04:08:36 Join bluebrother [0] (
04:08:36 Quit bluebrother (Changing host)
04:08:36 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
04:09:37 Quit InsDel (Read error: Connection reset by peer)
04:14:20 Quit AlexP (Remote host closed the connection)
04:14:22 Join Barahir_ [0] (
04:17:36 Quit Barahir (Ping timeout: 252 seconds)
04:18:02 Join AlexP [0] (~alex@rockbox/staff/AlexP)
04:19:34TheSevenkugel: not every DMAC supports mem2mem and per2per
04:19:55TheSeventhe nano2g DMAC officially doesn't support either one
04:20:08TheSevenhowever, there are some undocumented regs that look suspicious :)
04:30:15 Quit fdinel (Quit: Miranda IM! Smaller, Faster, Easier.
04:43:34 Quit Llorean (Quit: Leaving.)
04:46:05 Quit dys (Ping timeout: 276 seconds)
04:46:24 Join dys [0] (
04:49:31 Quit TheSeven (Ping timeout: 245 seconds)
04:50:30 Join Llorean [0] (
04:50:34 Quit Llorean (Changing host)
04:50:34 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
04:53:52***Saving seen data "./dancer.seen"
04:54:20 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
04:56:39 Quit amiconn (Disconnected by services)
04:56:39 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:56:57 Quit pixelma (Disconnected by services)
04:56:59 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:57:00 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:57:02 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
05:15:04 Join Gnea [0] (~gnea@unaffiliated/gnea)
05:22:51 Quit anewuser (Ping timeout: 245 seconds)
05:24:59 Join anewuser [0] (anewuser@unaffiliated/anewuser)
05:34:41 Quit ps-auxw (Ping timeout: 276 seconds)
05:35:33 Quit yosafbridge (Ping timeout: 245 seconds)
05:38:58 Quit MethoS- (Remote host closed the connection)
05:45:18 Join ps-auxw [0] (
05:47:46 Quit fyrestorm (Read error: Connection reset by peer)
06:07:46 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
06:12:13 Quit JesusFreak316 (Ping timeout: 245 seconds)
06:14:46 Quit GeekShadow (Quit: The cake is a lie !)
06:25:59 Join FM|Dostoevsky [0] (~Philip@
06:28:25 Quit Sundiver (Ping timeout: 245 seconds)
06:36:14 Quit anewuser ()
06:45:16 Join ce_CANTIK_cR_YG_ [0] (~BOLERO`
06:45:16 Part ce_CANTIK_cR_YG_ ("")
06:53:54***Saving seen data "./dancer.seen"
07:06:05 Quit panni_ (Read error: Connection reset by peer)
07:09:44 Join Horschti [0] (~Horschti@xbmc/user/horscht)
07:13:29 Quit Horscht (Ping timeout: 276 seconds)
07:19:51 Quit FM|Dostoevsky (Quit: Leaving)
07:20:52 Quit ReimuHakurei_ (Quit: Leaving)
07:21:59 Join ReimuHakurei [0] (~reimu@
07:22:57 Join froggyman [0] (~seth@unaffiliated/froggyman)
07:24:44 Join ReimuHakurei_ [0] (~reimu@
07:24:45 Quit ReimuHakurei (Read error: Connection reset by peer)
07:43:03 Quit timccc (Ping timeout: 245 seconds)
07:43:16 Join timccc [0] (~timccc@
07:46:30 Quit simonrvn (Read error: No route to host)
07:55:33 Join factor [0] (
07:56:07 Quit factor (Remote host closed the connection)
07:56:22 Join factor [0] (
08:08:17 Quit CaptainKewl (Ping timeout: 245 seconds)
08:13:13 Nick Loto_ is now known as Loto (
08:13:24 Quit Loto (Changing host)
08:13:24 Join Loto [0] (~Loto@xbmc/user/Loto)
08:18:42 Join eWill [0] (
08:19:49eWillnewegg has recertified Sansa Fuze v2 4GB for $30. Am I right that it doesn't matter what firmware version it comes with? (in order to install RB)
08:23:47 Quit Llorean (Quit: Leaving.)
08:24:53 Join simonrvn [0] (
08:25:30 Join Llorean [0] (
08:25:38 Quit Llorean (Changing host)
08:25:38 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
08:33:35 Join Llorean1 [0] (
08:33:48 Quit Llorean (Disconnected by services)
08:33:52 Nick Llorean1 is now known as Llorean (
08:33:54 Quit Llorean (Changing host)
08:33:54 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
08:35:41 Join stoffel [0] (
08:45:30 Quit Llorean (Quit: Leaving.)
08:47:14 Join Llorean [0] (
08:47:20 Quit Llorean (Client Quit)
08:53:55***Saving seen data "./dancer.seen"
08:54:46 Quit BHSPitMonkey (Remote host closed the connection)
09:05:56 Join yosafbridge [0] (
09:06:36 Part toffe82
09:20:52 Join JdGord [0] (
09:24:12 Join kevku [0] (~kevku@2001:7d0:0:f000::135d)
09:25:32 Join Rob2223 [0] (
09:28:51 Quit Rob2222 (Ping timeout: 245 seconds)
09:32:03 Quit sasquatch (Quit: WeeChat 0.3.2)
09:32:28 Join sasquatch [0] (
09:33:24 Join Buschel [0] (
09:34:28Buschelkugel: your measurements of the alignment patch look amazing. what version of the patch did you apply? especially what version of the codec patch?
09:36:36Buschelwas it one of those where only mpc uses the new aligment attribute?
09:42:49 Quit shai (Ping timeout: 265 seconds)
09:46:18 Quit JdGord (Quit: Bye)
09:54:42 Quit stoffel (Ping timeout: 276 seconds)
10:00:30 Join bmbl [0] (
10:00:30 Quit bmbl (Changing host)
10:00:30 Join bmbl [0] (~bmbl@unaffiliated/bmbl)
10:02:40 Quit factor (Read error: Connection reset by peer)
10:05:07 Join timccc1 [0] (~timccc@
10:06:33 Quit timccc (Ping timeout: 276 seconds)
10:19:42 Join factor [0] (
10:35:50 Join yzflcyq [0] (
10:36:18 Quit JdGordon (Ping timeout: 276 seconds)
10:37:44 Join ender` [0] (
10:39:46 Quit yzflcyq (Client Quit)
10:41:20 Join PurlingNayuki [0] (~PurlingNa@
10:47:24PurlingNayukiI think VX747 can be moved to stable target.
10:47:59 Part PurlingNayuki
10:48:55 Join bertrik [0] (
10:48:55 Quit bertrik (Changing host)
10:48:55 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
10:49:11 Join PurlingNayuki [0] (~PurlingNa@
10:50:28 Join DerPapst [0] (
10:51:16 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
10:51:41 Join {phoenix} [0] (
10:53:57***Saving seen data "./dancer.seen"
10:56:20 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
11:06:45 Part PurlingNayuki
11:08:03 Join PurlingNayuki [0] (~PurlingNa@
11:08:56 Part PurlingNayuki
11:12:07 Join PurlingNayuki [0] (~PurlingNa@
11:12:18 Part PurlingNayuki
11:15:07 Join PurlingNayuki [0] (~PurlingNa@
11:15:13 Part PurlingNayuki
11:17:04 Join PurlingNayuki [0] (~PurlingNa@
11:17:58 Join yzflcyq [0] (
11:19:55 Part PurlingNayuki
11:19:56 Quit yzflcyq (Client Quit)
11:20:40 Join Kupop [0] (
11:33:59 Quit eWill (Quit: ChatZilla 0.9.86 [Firefox 3.6.12/20101026210630])
11:49:41 Quit jhMikeS (Read error: Connection reset by peer)
11:49:59 Join jhMikeS [0] (~jethead71@rockbox/developer/jhMikeS)
11:57:28robin0800any news on the alternate lines viewport problem?
12:05:48 Quit bertrik (Read error: Connection reset by peer)
12:07:33 Quit robin0800 (Remote host closed the connection)
12:07:56 Join robin0800 [0] (
12:08:01 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
12:12:58 Join stoffel [0] (
12:24:43 Join teru [0] (
12:34:00 Quit stoffel (Ping timeout: 255 seconds)
12:44:45*bertrik tries to figure out stuff about the .sb file format
12:51:40 Quit tchan (Read error: Connection reset by peer)
12:52:30 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
12:52:40 Join TheLemonMan [0] (~lemonboy@
12:54:00***Saving seen data "./dancer.seen"
12:56:12 Quit liar (Ping timeout: 255 seconds)
12:57:24 Join liar [0] (
13:12:23 Quit user890104 (Ping timeout: 272 seconds)
13:18:15 Quit JdGordon (Ping timeout: 255 seconds)
13:22:16 Join MethoS- [0] (~clemens@
13:27:08CIA-7New commit by Buschel (r28625): Change alignment macros to allow further performance optimization. Define the CACHEALIGN macros for all ARM CPUs, the used alignment size is derived ...
13:32:28 Quit antil33t (Ping timeout: 245 seconds)
13:34:01Buschelhmm, is something wrong with the builds?
13:40:20 Join antil33t [0] (
13:41:58 Join xxcv [0] (
13:42:03 Quit Sudos (Ping timeout: 264 seconds)
13:42:33 Quit Kupop (Ping timeout: 255 seconds)
13:44:00 Join dfkt [0] (dfkt@unaffiliated/dfkt)
13:44:33 Join GeekShadow [0] (
13:44:34 Quit GeekShadow (Changing host)
13:44:34 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
13:45:08 Quit rasher (Read error: Connection reset by peer)
13:48:29 Quit n17ikh (Ping timeout: 272 seconds)
13:50:04 Nick YPSY is now known as Ypsy (
13:54:29 Quit MethoS- (Remote host closed the connection)
13:56:18 Quit benedikt93 (Ping timeout: 240 seconds)
13:57:36*Buschel kicks CIA-7
13:59:53 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
14:01:53bertrikit looks like the build didn't finish if you look at the build page
14:02:39Buschelyep, 33m is far too much. any idea what to do?
14:03:15BuschelI am not familiar with the build system....
14:03:22gevaertswait for Bagder or Zagor to kick it
14:03:54bertrikthis is just more likely to happen on week-ends :)
14:05:30bertrikWhich targets have an ARM11 CPU?
14:06:59amiconnThe beast (and some androids I presume)
14:08:47 Quit tchan (Read error: Connection reset by peer)
14:09:37 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
14:09:43Buscheltests on some arm9 with 32 byte cache alignment would also be interesting (tcc and as3525)
14:10:04 Join pamaury [0] (
14:10:05 Quit pamaury (Changing host)
14:10:05 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
14:16:06 Nick Horschti is now known as Horscht (~Horschti@xbmc/user/horscht)
14:25:13 Join Kupop [0] (
14:27:41 Join rasher [0] (
14:27:42 Quit rasher (Changing host)
14:27:42 Join rasher [0] (~rasher@rockbox/developer/rasher)
14:28:00 Quit liar (Ping timeout: 255 seconds)
14:28:40 Join liar [0] (
14:37:31 Quit rasher (Read error: Connection reset by peer)
14:37:59 Join n1s [0] (~n1s@rockbox/developer/n1s)
14:38:46 Join rasher [0] (
14:38:46 Quit rasher (Changing host)
14:38:46 Join rasher [0] (~rasher@rockbox/developer/rasher)
14:47:59CIA-7New commit by teru (r28626): FS #6321: Universal Image Viewer ...
14:49:49CIA-7r28626 build result: All green
14:54:03 Quit Gnea (Ping timeout: 240 seconds)
14:54:05***Saving seen data "./dancer.seen"
15:03:57CIA-7New commit by Buschel (r28627): Use MEM_ALIGN_ATTR in mpc codec. Speed up by 22-24% was measured on ARM11.
15:05:49CIA-7r28627 build result: All green
15:05:55 Quit antil33t (Read error: Connection reset by peer)
15:06:04 Join antil33t [0] (
15:08:45 Quit factor (Read error: Connection reset by peer)
15:09:47 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
15:16:52 Join fdinel [0] (
15:17:32 Join factor [0] (
15:18:22CIA-7New commit by moos (r28628): Bring the spanish translation up to date. ...
15:18:54*bertrik is making a tiny bit of sense of the .sb file format
15:19:19pamauryis it documented ?
15:20:16CIA-7r28628 build result: All green
15:20:22Buschelhmm, the use of CACHEALIGN_SIZE in our *.lds seems strange to me...
15:20:27bertriknot as far as I know
15:20:33 Quit noamsml (Ping timeout: 240 seconds)
15:20:52pamauryhow do you make sense of it then ?
15:20:53bertrikI do have a binary to create a .sb file, so I'm creating an empty file now and comparing it with an actual .sb file
15:21:06pamauryis it encrypted ?
15:21:19Buschelin some (e.g. PP) we use _SIZE 16, for some CPUs with the same size we use _SIZE 0x4 (e.g. S5L870x)
15:22:09bertrikpamaury, not the entire file is encrypted, some parts are plain (like some graphics), I think only the actual binary executable code is encrypted
15:22:23Buschelseems like a mixture of CACHEALIGN_SIZE and CACHEALIGN_BITS
15:22:23pamaurylast time I worked with sb files (for old STMPs), the format was 1) encrypted 2) a list of commands of the form WRITE <data> AT ADDRESS <addr>
15:22:28 Join noamsml [0] (~noamsml@
15:23:05bertrikI think I see some kind of timestamp in the file, a total file length, and the names/addresses/lengths of the sub-parts
15:23:33bertrikso, similar to our clipsplit tool, I want to make something like sbsplit
15:23:34pamaurywhat is the input format of that binary ? elf ?
15:24:45bertrikThere is a command file which specifies the parts to include, I'm now using an empty list of files for testing
15:25:39pamauryis it freely available ? I can try
15:27:15Buschelwe could define CACHEALIGN_SIZE in the firmware/export/<cpu>.h as well and use it in the *.lds then −− imx31 does it this way... not use whether the fixed numbers (like ALIGN(0x4) for S5L870x) were intentional... anyone?
15:27:33Buschel*not sure
15:31:22CIA-7New commit by teru (r28629): fix −−rbdir is ignored in configure
15:32:25 Join Gnea [0] (~gnea@unaffiliated/gnea)
15:33:21CIA-7r28629 build result: All green
15:34:15pamaurybertrik: can I help you and try to hack understand this format ?
15:34:36 Quit liar (Ping timeout: 255 seconds)
15:34:57bertrikmaybe, here's elftosb binary :
15:35:20 Quit timccc1 (Quit: Leaving.)
15:36:51CIA-7New commit by amiconn (r28630): Fix pointer aliasing warning which for some reason was only triggered in SH builds.
15:37:29bertrikyay, identified a microsecond timestamp
15:38:26pamaurywhat is the command file format ?
15:38:36CIA-7r28630 build result: All green
15:42:34bertrikpamaury, this thread has an example of the command file format:
15:43:05 Join liar [0] (
15:45:55pixelmateru: the new imageviewer.tex doesn't contain some warnings and notes from the old specific viewer descriptions, I think it should (e.g. about stopping playback, file size restrictions, jpeg format restrictions)
15:46:48pixelmaor it does
15:46:55pixelmaok, I read it again
15:48:06pixelmaI'm sorry I see now that it does
15:50:34terupixelma: btw, could you please test FS #11725?
15:51:03 Join n17ikh [0] (
15:53:53pixelmaprobably not today
16:01:18 Quit teru (Quit: Quit)
16:01:30 Quit Gnea (Ping timeout: 272 seconds)
16:10:15 Join anewuser [0] (anewuser@unaffiliated/anewuser)
16:19:07 Join Lear [0] (chatzilla@rockbox/developer/lear)
16:19:15n1samiconn: that aliasing warning seems strange, afaiu signed/unsigned pointers are allowed to alias each other, maybe a gcc4.0 bug?
16:26:36 Join Boldfilter [0] (
16:27:37CIA-7New commit by learman (r28631): Backport fix for FS #11696 to 3.7 branch: Scrollwheel doesn't respond in some cases.
16:32:07 Quit GeekShadow (Ping timeout: 260 seconds)
16:33:13 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
16:35:50*n1s can't decide what to do with this gcc upgrade for coldfire
16:36:34n1sI want to find out why it's slowing down vorbis decoding but profiling seems to not work so well with the new gcc, or at least the results are strange
16:38:37n1sAlso for profiling to work at all a nasty hack is required, which i don't see a clean way around but best is probably to write all of switch_thread in asm but that's annoying
16:40:53 Join balintx [0] (
16:43:01 Join CaptainKewl [0] (
16:45:45 Join shai [0] (
16:54:08***Saving seen data "./dancer.seen"
16:55:37 Quit CaptainKewl (Remote host closed the connection)
16:58:55 Quit markun (Read error: Connection reset by peer)
16:59:29 Join user890104 [0] (
17:09:03 Join markun [0] (~markun@rockbox/developer/markun)
17:09:23 Join panni_ [0] (
17:13:30Buscheln1s: can this vorbis slowdown be connected to alignment?
17:13:50Buscheln1s: wasn't it movem in DRAM which would be affected?
17:14:11n1sBuschel: it is possible, i dunno
17:16:35Buschelyou could simply try by aligning those data arrays that are used by the vect_() functions (e.g. vect_mult_fw())
17:16:38n1si also don't know how to figure it out, libtremor uses malloc for allocations so i would need to hack on that i assume
17:17:07Buschelhmm, wait
17:17:58n1salso the coldfire vector asm in tremor aligns to 16 bytes before using movem so the diff shouldn't bee this big
17:18:56n1sbut the 2MHz diff actually isn't too big a deal since most other codecs are either as fast as with 3.4 or faster i'm more interested in getting profiling working reliably
17:23:19n1si'm looking at the asm of the __cyg_profile_func_enter from the codeclib atm but can't understand what it's doing and why so i'll probably leave it alone for now
17:27:31n1sit looks like it passes an *instruction* as an argument to the core profiling functions...
17:29:45n1smoveal %sp@,%a0; movel %a0@(4),%sp@-;
17:30:00n1sis the first thing __cyg_profile_func_enter does
17:30:19 Quit Horscht (Quit: Verlassend)
17:31:07LearIsn't that the return address on the stack?
17:31:36 Join kugel [0] (~kugel@rockbox/developer/kugel)
17:31:37n1sLear: yes but IIUC it dereferences the return adress with an offset of 4 on the second instruction
17:31:59n1sand then pushes whatever that points to onto the stack ???
17:32:17n1s(as an argument for the core function)
17:38:08LearLooks like it wants the return address for the function that called __cyg_profile_func_enter. Did similar tricks on the Amiga...
17:38:31 Quit froggyman (Quit: Bye)
17:38:44pixelmaI wonder a bit about what to do with - it replaced the plugin actions that were used back then in the clock plugin (including "overlapping" contexts) with core actions. This has become obsolete (?) with the plugin action rework so only one context is used now and I'm not aware of any problems with the controls in the clock plugin. I'd close it (I also think it won't apply anymore) but am not sure if I'm
17:38:44pixelmaoverlooking something
17:39:17pixelmaother opinions?
17:39:18LearWhen it calls the core function, doesn't it include any more parameters?
17:40:04n1sLear: yes, it includes one that's passed to it
17:40:54kugelBuschel: I used the first two patches (v03 was already in svn at this point)
17:41:23 Quit xxcv (Ping timeout: 240 seconds)
17:41:37n1sLear: i'm looking at it since when building with gcc4.4 the profiling doesn't work
17:42:06LearOn Coldfire only?
17:42:37n1si don't remember if it works on arm or if i've tested it
17:43:01 Quit krazykit (Read error: Operation timed out)
17:44:57 Join Horscht [0] (
17:44:57 Quit Horscht (Changing host)
17:44:57 Join Horscht [0] (~Horschti@xbmc/user/horscht)
17:46:28pixelmaI'd also propose for closing - showing these settings should now be possible with the settings tag and the author himself seems to have lost interest since (and theme code has changed a lot too)
17:46:48 Join krazykit [0] (
17:47:08LearBy the way, that asm code you mentioned is probably emitted by "__builtin_return_address(1)", and looks reasonable to me.
17:47:58kugelpixelma: I agree, to closing both patches
17:48:59n1sLear: maybe i'm dense but since sp points to the return adress and that is loaded into a0, doesn't movel %a0@(4),%sp@-; *dereference* the return adress (as in push code onto the stack)?
17:49:03 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
17:49:54pixelmakugel: thanks, if no-one objects in the next hours, I'll close them
17:52:15 Quit fdinel (Ping timeout: 245 seconds)
17:52:25LearHm, not sure about that syntax. Is "@" enough to dereference?
17:53:13n1si think so, i'm not familiar with this syntax either but don't think objdump let's you output the other kind
17:54:28n1si think i've spotted the problem with gcc 4.4 btw, i'll post the asm in pastebin
17:55:01 Quit ender` (Quit: If you're going to do something tonight that you'll be sorry for tomorrow morning, sleep late.)
17:56:55n1sgcc 4.4 pushes the contents of a6 on the stack instead, which is the frame pointer, which we don't use...
17:59:35 Join stoffel [0] (
18:01:08n1salso i don't see why coldfire uses the __builtin_return_address() call
18:02:27 Join ender` [0] (
18:02:33n1si wonder if it's worth submitting another gcc bug report...
18:03:05 Quit benedikt93 (Ping timeout: 245 seconds)
18:03:05 Quit ender` (Read error: Connection reset by peer)
18:03:19 Join dfkt_ [0] (~dfkt@unaffiliated/dfkt)
18:03:33 Join ender` [0] (
18:05:54 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
18:05:59 Join mt [0] (~mt@rockbox/developer/mt)
18:06:50 Quit dfkt (Ping timeout: 245 seconds)
18:07:07 Quit TheLemonMan (Quit: Help me, i got shot! *DIES*)
18:08:59LearLooks like that could be the problem, yes. But I seem to remember reading somewhere that just submitting a report isn't enough. You'll need to post about it on the mailing list to get any attention to it.
18:09:17n1shmm, yes it seems __builtin_return_address always use the frame pointer so it is broken when building with -fomit-frame-pointer
18:09:33n1sLear: that or cc the right people in the bug report
18:10:38n1si've posted two coldfire specific bugs, one sat in their bugzilla for a year before i mailed some people i had seen touching coldfire stuff, the other got fixed in a few days when i cc'ed those same people in the bugreport
18:12:07 Quit krazykit (Ping timeout: 255 seconds)
18:13:37n1sbut i also don't see why we use the builtin instead of just the argument passed to the function
18:14:45LearBut the passed function is __builtin_return_address(0), isn't it?
18:16:54Buschelkugel: it is interesting too see a speed up on codecs the patch does not directly touch (e.g. aac). are those results reproducible? maybe a general speed up outside the codec (pcm/dsp/bitstream buffer) happened?
18:17:57kugelyour patches enabled CHACHEALIGN_ATTR, so every place which is already using it now without your patch could get faster
18:18:20n1sLear: AFAICT the functions pass their own start address as this_fn (first arg) and their return adress as call_site (second arg)
18:18:28Buschelkugel: I expect there is a lot potential for most codecs as mpc scaled that much via just aligning some of the arrays
18:18:38n1sbut it looks like gcc 3.4 is doing something wrong...
18:18:47n1sso that may be why we don't use this
18:19:10 Join krazykit [0] (
18:19:16 Quit Kupop (Ping timeout: 276 seconds)
18:19:21kugelBuschel: IIRC you sped up mpc in the meantime (the last ASM but non-alignment benches are from october)
18:21:16Buschelahh, ok. didn't know that...
18:22:23n1sLear: yes, comparing gcc 3.4 and 4.4 asm 3.4 is passing sp+4 as call_site which is wrong and i assume why we don't use it, 4.4 seems to have fixed this
18:23:15n1sso using the arguments should be fine on coldfire with gcc 4.4 then
18:23:52 Quit simonrvn (Read error: Connection reset by peer)
18:24:26 Join Kitr88 [0] (
18:25:55Buschelkugel: same has been done with libtremor a few days ago. would be interesting to see a comparison against the same, but unpatched, version
18:28:12 Quit Kitar|st (Ping timeout: 260 seconds)
18:28:17 Quit Lear (Quit: ChatZilla 0.9.86 [Firefox 4.0b8pre/20101120031924])
18:28:30 Quit robin0800 (Remote host closed the connection)
18:31:54 Join piomurava [0] (
18:33:47piomuravai've a problem building
18:33:48 Quit krazykit (Quit: hang on)
18:34:02piomuravawith cygwin
18:34:24piomuravai've searched eveywhere
18:34:31piomuravabut i can't solve!
18:34:34 Join krazykit [0] (
18:37:02piomuravawhen i ask to "make"
18:37:09piomuravai have a problem
18:38:07 Quit piomurava (Quit: CGI:IRC (EOF))
18:38:16 Join domonoky1 [0] (
18:38:23 Join piomurava [0] (
18:39:03piomurava$ make CC tools/rdf2binary /rockbox/tools/rdf2binary.c:26:23: error: sys/types.h: No such file or directory
18:39:20 Quit domonoky (Ping timeout: 245 seconds)
18:39:36piomuravawhat i'm doing wrong?
18:40:42n1spiomurava: sounds like you are missing some system headers package
18:42:28piomuravahow i cancheck?
18:42:40n1sdid you follow the guide on the wiki to install?
18:42:53piomuravaall the guide
18:43:03piomuravaall the package
18:43:06n1swell, check that you completed all the steps maybe?
18:43:45piomuravamm yes... i have check twicw
18:49:16piomuravai had some problems installing crosscompiler
18:49:31piomuravaarm eai
18:49:35piomuravaarm eabi
18:50:17piomuravanow it say
18:50:19piomuravaUsing arm-elf-eabi-gcc 4.4.4 (404)
18:50:28piomuravacoulb be a problem?
18:50:44n1swhy would that be a problem?
18:54:09***Saving seen data "./dancer.seen"
18:54:36 Join piomurav [0] (
18:54:41n1shmm, i now have profiles from both 4.4 and 3.4 both have the *exact* same number of total calls but differ widly in the number of calls per function and also in which functions are inclued at all!?
18:55:22 Quit stoffel (Remote host closed the connection)
18:55:49 Quit piomurava (Quit: CGI:IRC (Ping timeout))
18:56:57kugeln1s: perhaps different inlining strategies?
18:58:45CIA-7New commit by Buschel (r28632): Use MEM_ALIGN_ATTR in libdemac instead of fixed alignment. Speeds up arm11 by ~6%.
18:59:18piomuravwhat i can do?
19:00:00n1skugel: that's probably part of it but still it looks strange
19:00:14 Join CaptainKewl [0] (
19:00:35CIA-7r28632 build result: All green
19:01:33n1ssome functions are inlined by 4.4 and thus their entries aren't counted but the fact that the total count is identical makes me wonder
19:02:34piomuravmy problem could depend from this?
19:05:18n1spiomurav: no that was not directed at you, i've not messed with cygwin in a long time so i don't know much of what could be wrong except something missing
19:05:43n1sif you hang around maybe someone who knows will drop in
19:06:33piomuravok thank
19:07:42Buschelkugel: can you give this patch a try? it changes the alignment of oggmalloc iram buffers:
19:08:42kugelI'll try this week, no time today
19:09:11 Join slater [0] (
19:09:47Buschela pitty... :/
19:10:32Buschelanyone else with an arm11 target (e.g. beast) having time for a test?
19:11:50slaterhello everybody. I am having some troubles with libsdl on my player, libsdl seems to work, I can start simple loopwave testprogram but for video its giving me : Couldn't initialize SDL: No available video device even though I set SDL to use my /dev/fb0 any ideas?
19:13:15n1sah, "calls" to inlined functions were counted as calls to the parent function, that explains part of that...
19:13:17gevaertsslater: I suspect asking the sdl people might be more successful
19:14:23 Quit piomurav (Quit: CGI:IRC (EOF))
19:17:05Dreamxtremequick question
19:17:25 Join Kupop [0] (
19:17:46Dreamxtremeis there any port that will be close to the classic port when its ready lol so i can look at some documentation and learn rockbox
19:18:19kugelBuschel: doesn't apply
19:18:31 Quit guymann (Ping timeout: 265 seconds)
19:18:49n1sDreamxtreme: the interface will behave the same as for other ipods, i expect
19:19:07 Join guymann_ [0] (~charles@
19:19:19Dreamxtremeo ok
19:19:24Dreamxtremeso like the Nano
19:19:38Dreamxtremebut i'd assume for advance right?
19:19:42n1sBuschel: i made a patch long ago that drops usage of that whole iram malloc thing on targets that don't have malloc but it had a subtle bug that i never worked out
19:19:45kugelBuschel: nvm, it conflicted with your other patch
19:19:45Dreamxtrememore even
19:20:36n1sstripwax said he was going to look at it since he wrote the iram malloc stuff but he never did
19:21:04Buschelkugel: can you just restore svn before applying this patch (only libtremor/oggmalloc.c)?
19:21:15kugelBuschel: oh I still have your patches applied and can't get to svn right now. it needs to wait
19:21:20gevaertsDreamxtreme: I'd assume the video will be more similar to the classic than the nano from a user point of view
19:21:22n1sfs#11268 if anyone cares
19:21:57kugelwait, I can s/MEM_ALIGN_ATTR/CACHEALIGN_ATTR/ for this file
19:22:47Buschelkugel: the patch could be applied very simple "by hand", if you would take a look at that patch. only if you find time, of course
19:24:18Dreamxtremeoooo ok
19:24:32*Buschel is skimming through lib/tlsf/src/tlsf.c
19:24:34Dreamxtremestill you can do more with a classic
19:24:44Dreamxtremeim seriously looking forward to rockbox
19:24:47Dreamxtremeon a classic
19:24:50Buschelis this the malloc implementation for ogg?
19:24:56gevaertsDo more? What sort of thing?
19:24:59n1sBuschel: yes
19:26:33Buscheln1s: there is an alignment defined in it as well. it is set to 2*sizeof(void*) = 8 byte
19:27:02n1sBuschel: ah, that might benefit us then
19:27:28 Join toffe82 [0] (
19:27:38Buscheln1s: wouldn't this be worth a quick test on CF or arm11 as well? ;o) -> "#define BLOCK_ALIGN (16)"
19:28:00Buscheln1s: could you perform a short test on your CF target?
19:28:23n1ssure, a quick test
19:28:53Buscheln1s: if it works I can search for a good solution that might use CPU dependent alignment sizes
19:28:55 Quit Loto ()
19:29:56n1sBuschel: what should i change?
19:30:36Buschelkugel: thank you! what is the cpu clock?
19:31:01kugelBuschel: 600MHz
19:31:06kugelno real difference from that patch
19:31:29Buscheln1s: lib/tlsf/src/tlsf.c, line 127 -> #define BLOCK_ALIGN (16)
19:33:55Buschelkugel: yep, seems like the allocated buffer sizes are multiples of 32 bytes :)
19:34:12Buschelat least it won
19:34:16Buschel't harm
19:36:07kugelwell it does harm if it makes small allocations unnecessarily big
19:37:01kugel(a tiny bit)
19:39:29 Join Lear [0] (chatzilla@rockbox/developer/lear)
19:40:22 Quit kugel (Remote host closed the connection)
19:40:29n1sBuschel: it doesn't seem to make any difference on CF
19:40:39 Join T44 [0] (
19:41:18Buscheln1s: ok :/
19:44:20 Quit Topy (Ping timeout: 245 seconds)
19:44:24 Join JesusFreak316 [0] (
19:48:49CIA-7New commit by Buschel (r28633): Remove unused define.
19:49:27 Join InsDel [0] (
19:50:01 Join hkmix [0] (
19:50:42CIA-7r28633 build result: All green
19:53:27 Quit n1s (Quit: Lämnar)
19:54:53 Part slater
19:56:49 Join AMS2 [0] (
19:58:01 Quit AMS2 (Client Quit)
19:59:27CIA-7New commit by Buschel (r28634): Comment unused function.
20:00:04 Join amsv2 [0] (
20:01:17 Quit amsv2 (Client Quit)
20:01:30CIA-7r28634 build result: All green
20:16:18 Join stripwax [0] (
20:17:22 Join Gnea [0] (~gnea@unaffiliated/gnea)
20:17:42 Join stripwax__ [0] (
20:29:56 Quit InsDel (Ping timeout: 276 seconds)
20:33:36 Nick dfkt_ is now known as dfkt (~dfkt@unaffiliated/dfkt)
20:33:51 Join evilnick [0] (~evilnick@rockbox/staff/evilnick)
20:39:08 Join simonrvn [0] (
20:42:43 Quit Gnea (Ping timeout: 272 seconds)
20:47:37 Join Gnea [0] (~gnea@unaffiliated/gnea)
20:48:46 Quit GeekShadow (Ping timeout: 252 seconds)
20:49:14 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
20:54:10***Saving seen data "./dancer.seen"
20:55:42 Join iq [0] (~iq@unaffiliated/iq)
20:59:12 Quit AlexP (Remote host closed the connection)
21:06:08 Join AlexP [0] (~alex@rockbox/staff/AlexP)
21:06:31 Quit stripwax (Quit:
21:10:47 Quit stripwax__ (Read error: Connection reset by peer)
21:11:28 Join InsDel [0] (
21:14:06 Join AMSv2 [0] (
21:14:22 Join stripwax [0] (
21:15:18 Quit AMSv2 (Client Quit)
21:16:40 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
21:19:44 Quit hkmix (Quit: CGI:IRC)
21:21:35bertrikpamaury, I have a really basic .sb parser right now, it can extract parts but not decrypt them
21:22:32 Quit GeekShadow (Quit: The cake is a lie !)
21:23:09pamauryI guess to decrypt them you'll have to reverse-engineer the program. And actually, is the program using the same encryption key as the processor ?
21:24:10bertrikI suppose so
21:24:46bertrikthe elftosb tool also allows use of a "zero-key", with some luck they didn't bother to actually pick a real key and used some default one
21:25:23 Quit Gnea (Ping timeout: 240 seconds)
21:25:24pamauryare you sure or is it a guess ?
21:26:01bertrikI read somewhere in an iMX document that a session key generated by elftosb is also involved, so some parts of the file are different every time
21:26:06bertrika guess of course
21:27:05pamauryWell, to be sure, one would have to know the format and the encryption algorithm
21:28:31 Quit ReimuHakurei_ (Ping timeout: 245 seconds)
21:28:53pamauryAnyway, it will be necessary to find out which algorithm it is to decrypt the firmware no ?
21:30:40pamauryhum, so one will have to reverse engineer this program :(
21:32:46bertrikOr maybe a deeper google search, or search through imx233 documents about clues
21:33:13bertrikI think I remember seeing hashing or decrypting hardware in a block diagram somewhere today
21:35:12pamauryyes but it doesn't mean you have the algorithm. In the case of the STMP I worked on, the algorithm was not publicily available and I had to find out what it was. Hopefully, it was trivial but it's probably not the case here
21:35:29bertrikThe imx233 document linked from the SansaFuzePlus wiki page shows 128-bit AES and SHA-1 blocks
21:37:10pamauryactually the document even says: Secure Boot 128-bit AES hardware decryption
21:37:12bertrikThere's an I2C EEPROM in the sansafuzeplus that could be read out
21:38:06pamaurywould it be useful ?
21:39:01 Quit stripwax (Quit:
21:39:34bertrikI'd have to read it out first :)
21:41:02pamauryhehe, they find someone with a fuze+ and something to read out the eeprom
21:42:31 Quit Horscht (Quit: Verlassend)
21:48:26 Quit jordan` (Ping timeout: 255 seconds)
21:49:01 Join Horscht [0] (
21:49:01 Quit Horscht (Changing host)
21:49:01 Join Horscht [0] (~Horschti@xbmc/user/horscht)
21:49:46 Join Gnea [0] (~gnea@unaffiliated/gnea)
21:54:24 Join stripwax [0] (
21:58:32 Quit Lear (Quit: ChatZilla 0.9.86 [Firefox 4.0b8pre/20101121032004])
21:59:53 Join jordan` [0] (
22:01:51 Quit bmbl (Quit: Verlassend)
22:04:43 Quit factor (Ping timeout: 245 seconds)
22:11:33 Join factor [0] (
22:12:48 Quit stripwax (Ping timeout: 260 seconds)
22:13:14 Quit factor (Read error: Connection reset by peer)
22:13:43 Join factor [0] (
22:14:03 Join ReimuHakurei [0] (
22:16:25 Join stripwax [0] (
22:20:08 Quit Keripo (Read error: Connection reset by peer)
22:24:31 Join Keripo [0] (
22:27:30 Quit ReimuHakurei (Read error: Connection reset by peer)
22:28:04 Join ReimuHakurei [0] (~reimu@
22:30:58bertrikpamaury, this is the code I have to far to parse the .sb file:
22:31:38 Quit benedikt93 (Quit: Bye ;))
22:31:39bertrikand these are my notes about the format:
22:32:16 Quit ReimuHakurei (Ping timeout: 245 seconds)
22:35:22pamaurywhich chunks are encrypted ? Or suspected to be
22:36:17bertrikeverything but rsrc
22:42:38 Quit krazykit (Quit: awe yeeeeeee)
22:43:04 Join krazykit [0] (
22:44:30pamaurybertrik: do you have test example so I can just run elftosb2 on it ?
22:47:48 Quit stripwax (Quit:
22:50:05 Quit bertrik (Quit: :tiuQ)
22:50:24 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
22:51:36 Quit factor (Ping timeout: 240 seconds)
22:51:37pamaurythanks, in the command file, what the "load" and "call" commands are supposed to mean ?
22:53:27 Join factor [0] (
22:54:11***Saving seen data "./dancer.seen"
22:54:48CIA-7New commit by Buschel (r28635): Remove tabs.
22:56:40CIA-7r28635 build result: All green
23:00:16 Join froggyman [0] (~seth@unaffiliated/froggyman)
23:02:08CIA-7New commit by Buschel (r28636): Use MEM_ALIGN_ATTR in pitch detector plugin.
23:03:58CIA-7r28636 build result: All green
23:06:01 Quit kevku (Quit: KVIrc 4.0.2 Insomnia
23:06:41 Join stripwax [0] (
23:18:26 Quit Kupop (Ping timeout: 255 seconds)
23:21:23 Join einhirn [0] (
23:22:00 Quit einhirn (Client Quit)
23:22:10 Join einhirn [0] (
23:25:38 Quit Keripo (Read error: Connection reset by peer)
23:30:03 Join Jonas29 [0] (
23:30:37Jonas29hello I would like a information about your system
23:31:52 Quit {phoenix} (Remote host closed the connection)
23:32:06Jonas29do you exist a french support with another channel or here?
23:32:22 Quit aevin (Ping timeout: 240 seconds)
23:36:27 Quit factor (Read error: Connection reset by peer)
23:36:33 Quit domonoky1 (Read error: Connection reset by peer)
23:36:39 Join factor [0] (
23:38:56 Quit Buschel (Ping timeout: 276 seconds)
23:39:56 Quit einhirn (Read error: Connection reset by peer)
23:41:05 Quit factor (Read error: Connection reset by peer)
23:41:26 Join factor [0] (
23:51:47saratogadoes the beast only charge off of AC power?
23:55:55 Join JdGord [0] (
23:58:11 Quit pamaury (Remote host closed the connection)
23:59:01saratogais there someway to specify that struct members should be aligned?

Previous day | Next day