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 2013-12-26

00:05:20***Saving seen data "./dancer.seen"
00:06:01 Quit LordBimbam (Read error: Operation timed out)
00:08:18 Quit lebellium_gs2 (Quit: Bye)
00:32:46saratogaits incredible how bad i am at ARM programming
00:35:51coppertry coding with your feet?
00:36:06 Quit Narod ()
00:55:22 Quit JordiGH (Ping timeout: 260 seconds)
01:06:34 Quit ender` (Quit: If man could be crossed with the cat it would improve man, but it would deteriorate the cat. -- Mark Twain)
01:09:25 Join treaki__ [0] (
01:13:10 Quit treaki_ (Ping timeout: 252 seconds)
01:37:27saratogaanyone feel like looking at this pretty simple ASM and tell me what i'm missing?
01:37:38saratogathe output is just static, so i'm probably messing up the pointers ors omething
01:42:36 Quit bertrik (Remote host closed the connection)
01:53:20saratogahmm error on one of the constraints tot he asm function
02:05:22***Saving seen data "./dancer.seen"
02:07:22 Part rkbx_user
02:08:46derfsaratoga: BTW, you've seen jmspeex's recent commits?
02:08:54derfHe's been doing some MDCT work.
02:09:31saratogaderf: no i'll look now
02:10:59derfI think he's also considering reverting;a=commitdiff;h=f7547a4e and;a=commitdiff;h=72513f3cc
02:11:16derfWhich would cost 2400 bytes of table storage but appears to save about 3% decoding time.
02:11:28saratogaderf: thanks, folding in the bitreverse would slightly change my code
02:13:07saratogaderf: ok if hes going to revert that commit that would change quite a lot
02:14:01derfI think he's seeking opinions on whether or not that's a good idea.
02:15:21saratogado i understand correctly that that commit combines the different MDCT trig tables ?
02:15:44saratogaok we did that years ago for vorbis and ended up regretting it
02:16:07derfAny reason in particular?
02:16:11saratogait saves you something like half as much memory as your largest block size but ends up hurting you more on deeply pipelined processors
02:16:43saratogasince the memory accesses aren't as regular
02:17:10saratogawe found that it was a lot faster on arm7tdmi systems with SRAM, and slower on everything else
02:17:18derfWell, I would expect most of the data for Opus to fit in cache.
02:17:41saratogawell for us a deeply pipelined processor is ARM 11 with 16 KB cache :)
02:17:45derfI agree it's slower. As I said, he measured reverting it sped things up about 3%.
02:17:59derfThough I have no idea what he measured on (probably x86).
02:18:48saratogaas long as you mostly use the largest block size its not a big deal
02:19:08saratogai think we have problems with vorbis because the stride != 1 for the most common block size
02:19:31derfRight, with Opus the most common block size is the largest.
02:19:40derfUnless you're encoding fatboy or something.
02:19:42saratogaso you were loading lots of cache blocks and only using half the data in them
02:20:35saratogacache lines that is
02:21:05saratogaon arm7 it didn't matter because the cache is so useless anyway, but on arm11 it ended up being a big deal
02:21:38saratogawell i'll try and figure out why my code isn't working, and then adapt it to whatever he does
02:21:50saratogaprobably not too hard, i think 95% of my work is just trying to remember ARM asm
02:26:54derfThat's what QRC0001_UAL.pdf is for.
02:27:30saratogayeah but looking at the opcodes is not the same as remembering how to do things
02:37:51saratogafixed a couple stupid mistakes but it still doesn't work
02:38:33saratogaone more mistake
02:50:10saratogawhats the op to make a register negative?
02:50:56derfRSB rX,#0
02:51:38saratogaha yes of course
02:53:44 Join tertu [0] (
03:08:11 Quit ^7heo (Disconnected by services)
03:08:44 Join ^7heo [0] (
03:24:11 Join bzed_ [0] (
03:25:29 Join Kohlio [0] (
03:26:02 Join GeekShad1w [0] (
03:33:43 Join albb0920_ [0] (
03:33:56 Quit Kohlrabi (*.net *.split)
03:33:56 Quit GeekShadow (*.net *.split)
03:33:56 Quit Slasheri (*.net *.split)
03:33:56 Quit Ketturi (*.net *.split)
03:33:56 Quit albb0920 (*.net *.split)
03:33:56 Quit bzed (*.net *.split)
03:33:58 Nick bzed_ is now known as bzed (
03:38:53 Join Slasheri [0] (miipekk@rockbox/developer/Slasheri)
03:41:04 Join Ketturi [0] (
03:42:12 Quit soap (Ping timeout: 240 seconds)
03:43:25 Quit albb0920_ (Ping timeout: 240 seconds)
03:43:41 Join soap [0] (~soap@rockbox/staff/soap)
03:44:39 Join albb0920 [0] (
04:05:24***Saving seen data "./dancer.seen"
04:23:54 Quit amiconn (Disconnected by services)
04:23:54 Quit pixelma (Disconnected by services)
04:23:55 Join pixelma_ [0] (pixelma@rockbox/staff/pixelma)
04:23:55 Join amiconn_ [0] (amiconn@rockbox/developer/amiconn)
04:23:57 Nick pixelma_ is now known as pixelma (pixelma@rockbox/staff/pixelma)
04:23:59 Nick amiconn_ is now known as amiconn (amiconn@rockbox/developer/amiconn)
05:04:20saratogaah fixed it
05:18:59saratogaha, the hand written armv4 version is faster than the gcc version using the vastly better armv5 multiply instructions
05:19:05saratoganot bad
05:19:40saratoganow to try it in armv5
05:25:04 Quit TheSeven (Disconnected by services)
05:25:18 Join [7] [0] (~quassel@rockbox/developer/TheSeven)
06:04:38saratogaheres a working ARMv4 and ARMv5:
06:04:44saratogaARMv5 is not yet optimal
06:05:07saratogabut should be faster than armv4
06:05:27***Saving seen data "./dancer.seen"
06:13:05 Join mortalis [0] (~kvirc@
06:13:26saratogan1s_web: (for the logs) have you seen the opus fft changes the last few days?
06:13:58saratogai have some code that could go in our current git, but theres no point until we've synced to current opus-dev
06:35:03 Quit tchan (Ping timeout: 240 seconds)
06:35:52 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
06:54:44saratogaalso, whats the easiest way to pull a bunch of changes from the opus dev branch into rockbox? i'm assuming git has a workflow for this?
07:26:55 Quit tertu (Ping timeout: 260 seconds)
07:58:37 Quit cmhobbs (Ping timeout: 272 seconds)
08:05:28***Saving seen data "./dancer.seen"
08:53:16 Join kevku [0] (~kevku@2001:470:27:773:0:feed:c0f:fee)
10:05:31***No seen item changed, no save performed.
10:09:19 Join APLU [0] (~mulx@2a01:e34:ee29:12b0::10)
10:21:32 Join ender` [0] (
10:39:42 Join bertrik [0] (
10:39:42 Quit bertrik (Changing host)
10:39:42 Join bertrik [0] (~quassel@rockbox/developer/bertrik)
11:00:18 Join rela [0] (~x@pdpc/supporter/active/rela)
11:11:21 Join nls_web [0] (5ae0bbbe@gateway/web/freenode/ip.
11:13:35nls_websaratoga: yeah i just saw them, the merging of the bitrev reoirdering into the imdct was something i suggested a while ago and it helps a performance bit on platforms with slow memory but very little on PP
11:14:21nls_webhow much does your prerotate asm help?
11:18:03nls_webThe way i synced our codec with upstream was (IIRC) to copy over the files and cherry pick all the relevant commits in the codec dir but now that we're much closer to upstream i think just diffing the files against upstream 1.1 and applying that on the new file should be easy. Or i can send you a patch of all our changes when i get back to my desktop in two weeks or so
11:18:39nls_webProbably should have done the sync in two commits so that this would be easier...
11:23:42bertrikby the way, IIRC, when doing the initial port, we basically just added opus files until the linker no longer complained, there might be some code in our tree which is only useful for encoding
11:40:45 Quit akaWolf (Remote host closed the connection)
11:42:19 Join akaWolf [0] (~akaWolf@unaffiliated/akawolf)
11:51:41 Quit DexterLB (Read error: Connection reset by peer)
11:54:50 Join kugel [0] (
11:54:50 Quit kugel (Changing host)
11:54:50 Join kugel [0] (~kugel@rockbox/developer/kugel)
11:56:20 Join DexterLB [0] (~dex@
11:57:23kugelnls_web: that's equivalent to git rebase, so you can probably automate that
11:58:19nls_webkugel: probably but i don't know the git stuff that well
11:59:25nls_webbertrik: yeah, there's some code we don't use in there. I #Ifdef'ed out some but not everything
12:01:19nls_webkugel: how would you do it?
12:05:34***Saving seen data "./dancer.seen"
12:08:23 Join pikas [0] (512cceec@gateway/web/freenode/ip.
12:08:52pikassomeone could help me with an issue with ipod video and rockbox 3.13?
12:17:04nls_webpikas: ask and someone will answer if they know and when they see the question
12:20:29pikaswell, my problem is the next. i'm using rockbox since long ago, but this weekend, i connected my ipod in disk mode to my computer to copy some mp3, and when i restart my ipod, it gets freezy at rockbox welcome screen, and it doesnt continue
12:20:41pikasno error message or something else is shown
12:21:03pikassimply it doesn't pass the welcome screen
12:22:12pikasi tryed to copy rockbox folder of new, but the issue still persists
12:27:15 Quit bertrik (Quit: No Ping reply in 180 seconds.)
12:27:28 Join bertrik [0] (
12:27:28 Quit bertrik (Changing host)
12:27:28 Join bertrik [0] (~quassel@rockbox/developer/bertrik)
12:32:51 Quit kugel (Ping timeout: 272 seconds)
13:00:37gevaertspikas: I'd start out by checking for disk corruption
13:03:11pikasok, thanks! i would try chkdsk
13:03:26pikasor restoring ipod with itunes if there is not other way
13:05:29 Join ikeboy [0] (
13:09:33 Join kugel [0] (
13:09:33 Quit kugel (Changing host)
13:09:33 Join kugel [0] (~kugel@rockbox/developer/kugel)
13:15:30ikeboyCan I raise the number of bookmarks supported in the textviewer?
13:18:45gevaertsikeboy: you can change the number in the source and rebuild
13:19:27ikeboySo which line in which file would it be?
13:19:45gevaertsFrom what I can see, tv_bookmark.h
13:20:32pikasgood bye 2 all! and thanks for answering gevaerts ;)
13:20:32kugelnls_web: sorry, had to run. I know it's possible but I can't resemble the command right now
13:20:41 Quit pikas (Quit: Page closed)
13:21:32kugelindeed, the easiest might be to generate the diff to upstream at the last sync and apply that
13:22:46ikeboygevaerts: the line above the def says /* Maximum amount of register possible bookmarks */ might that be a hardware issue?
13:23:10gevaertsikeboy: why would it be? That comment just describes what the number means...
13:23:42ikeboyI'll try in the sim and see.
13:24:02gevaertsikeboy: if you read more comments, you'll notice that text_viewer wasn't written by someone who's very fluent in English, so some of these are a bit awkwardly worded
13:26:35 Quit rela (Read error: Connection reset by peer)
13:27:24ikeboyIn the sim build the textviewer isn't starting at all when I change the value to 32
13:28:39 Join lorenzo92 [0] (
13:30:10ikeboyI'll check the logs later I'm leaving
13:30:13 Quit ikeboy (Quit: ikeboy)
13:32:56lorenzo92anyone willing to review g#395? I don't want to bump, but i would like to see that committed to avoid continuous rebasing ^^
13:32:59fs-bluebotGerrit review #395 at : Samsung YP-R1 target port by Lorenzo Miori (changes/95/395/6)
13:35:11lorenzo92(i've spotted some whitespaces arghhh xD)
13:49:07bertriklorenzo92: is it mostly a copy of the yp-r0, or are there significant changes between them?
13:49:39lorenzo92yes, the big changes (to make the best code reuse) were already did some time ago...
13:49:48lorenzo92but I spotted a bug right now :D
13:50:13 Quit kugel (Ping timeout: 248 seconds)
13:51:32bertrikjust wondering about code duplication
13:53:24lorenzo92no there is no duplicate code
13:54:06bertrikcool, then I personally have no objection getting that committed
13:55:07lorenzo92because I have already prepared the "environment" with g#409
13:55:09fs-bluebotGerrit review #409 at : Samsung YP-R0 hosted target code refactoring by Lorenzo Miori (changes/09/409/3)
13:55:37lorenzo92bertrik: okay, good :) but now I realized that I still have to fix fm radio audio routing
13:56:16lorenzo92perhaps I should do that before committing?
13:56:29bertrikplease do if you know how
14:01:32bertrikbut it's not blocking to get the port accepted IMO
14:05:36***Saving seen data "./dancer.seen"
14:14:05 Join rela [0] (~x@pdpc/supporter/active/rela)
14:28:14 Join onder`_ [0] (
14:28:14 Quit DexterLB (Read error: Connection reset by peer)
14:30:01 Quit onder` (Ping timeout: 272 seconds)
14:30:04 Nick onder`_ is now known as onder` (
14:32:33 Join DexterLB [0] (~dex@
14:35:37 Join onder`_ [0] (
14:35:47 Quit onder` (Ping timeout: 261 seconds)
14:35:56 Nick onder`_ is now known as onder` (
14:52:03 Join amayer [0] (
14:55:05 Join Tukeke [0] (~Tukeke@unaffiliated/hlvsv)
15:01:22 Quit Tukeke (Quit: Tukeke)
15:19:03 Quit Cultist (Read error: Connection reset by peer)
15:19:07 Join Cultist_ [0] (~CultOfThe@2601:d:9280:fc:8634:97ff:fe17:5dc3)
15:20:10 Nick Cultist_ is now known as c (~CultOfThe@2601:d:9280:fc:8634:97ff:fe17:5dc3)
15:20:12 Nick c is now known as Cultist (~CultOfThe@2601:d:9280:fc:8634:97ff:fe17:5dc3)
15:24:12 Nick DormantBrain is now known as SuperBrainAK (~andy@2001:470:8:a61::5f92:59a1)
16:00:35 Quit mortalis (Quit: KVIrc 4.3.1 Aria
16:05:40***Saving seen data "./dancer.seen"
16:07:43 Quit Riviera (Quit: :))
16:14:33 Join shamus [0] (
16:15:22 Join cmhobbs [0] (
16:15:22 Quit cmhobbs (Changing host)
16:15:22 Join cmhobbs [0] (~cmhobbs@fsf/member/cmhobbs)
16:34:33 Quit rela (Read error: Connection reset by peer)
16:36:42 Join ikeboy [0] (
16:48:10 Quit nls_web (Ping timeout: 272 seconds)
16:55:33 Join y4n [0] (~y4n@unaffiliated/y4ndexx)
17:05:41lorenzo92bertrik: yeah got fm radio working, finally ^^
17:05:51lorenzo92it was a combination of problems actually
17:06:21lorenzo92now I have to fix also right channel xD
17:09:54 Quit Cultist (Read error: Connection reset by peer)
17:10:18 Join Cultist [0] (~CultOfThe@2601:d:9280:fc:8634:97ff:fe17:5dc3)
17:10:43 Join onder`_ [0] (
17:11:29bertriknice, what was the problem?
17:11:40bertrikwhat kind of radio chip is in there anyway?
17:12:26bertrikoh LV24020, didn't expect that
17:12:41bertrikargh, wait, SI4700 or compatible
17:12:47 Quit onder` (Ping timeout: 272 seconds)
17:12:50 Nick onder`_ is now known as onder` (
17:13:19lorenzo92so the problems were 2:
17:14:06lorenzo921) the call to audio_input_mux wasn't actually working (due to HOSTED protections in audio_path.c)
17:14:38lorenzo922) the codec was not correctly set (i.e. the aux mux was not enabled, due to different configuration w.r.t toshiba model(s))
17:15:55ikeboyTrying now with that bookmark change on a sim gives a: "Segmentation fault (core dumped)" error when textviewer is opened
17:15:59lorenzo92hum cabbiev2 radio screen does not support touchscreen unfortunately so i can't test seek (manual configuration via config.cfg is the only viable solution so far haha)
17:16:14lorenzo92but RDS and audio output are working fine :)
17:16:17 Join rela [0] (~x@pdpc/supporter/active/rela)
17:19:49bertrikI should add si4700 interrupt support some time
17:22:20lorenzo92hum you mean the RDS interrupt? Actually I'm doing software control, also on R0
17:22:41bertriklorenzo92: both for tuning and RDS
17:22:52lorenzo92ah okay
17:23:01bertriktuning can possibly be faster that way, getting an interrupt instead of waiting for the max tune time
17:23:36lorenzo92unfortunately on hosted targets setting an interrupt is quite tricky because of linux
17:24:03lorenzo92if not working on kernel side, of course...
17:24:07bertrikalso many tuners seem to support an "seek forward/backward" already, which we don't support yet on rockbox
17:24:50lorenzo92you mean a button that once triggered, does automatic scan until a good signal is found? in case, yea that could be very useful
17:25:55bertrikwe could add that, and use the HW support for tuners that support it and emulate it in software for tuners that don't
17:28:42bertrikDoes a sim make any sense for hosted targets?
17:32:53bertriklorenzo92: I can build a sim for yp-r0 (still compiling) but not for yp-r1
17:33:39lorenzo92bertrik: yes and no actually :) in the case of R0 and R1 you can't run on any linux.
17:34:00lorenzo92because we do much more than simply using kernel APIs (we actually control the HW)
17:34:14lorenzo92and yes, I fixed yp-r0 sim some time ago
17:34:26lorenzo92so i should also fix it for yp-r1
17:35:08lorenzo92and well, it makes sense to test out buttons / themes / more...especially on R1 that has a touchscreen
17:36:38 Quit rela (Ping timeout: 240 seconds)
17:37:15 Join Strife89 [0] (~Strife89@2602:306:250d:18b9:b130:d769:abf9:f661)
17:42:44bertriklorenzo92: I can't really understand what is going on in wmcodec_write, can you make it simpler?
17:42:47 Quit habys (Read error: Connection reset by peer)
17:43:31lorenzo92bertrik: yes, indeed, I don't actually remember why i did so ^^ perhaps to be sure of the correctness
17:43:38lorenzo92while developing
17:43:59lorenzo92i'll commit the latest modifications in a few
17:44:07 Join habys [0] (
17:49:51 Quit Cultist (Ping timeout: 240 seconds)
17:50:06saratoganls_web: the prerotate is a lot faster, but I think when I tested last year, pre and post rotate were each only about 10 or 15% of the total run time
17:50:40saratogaso probably a couple percent (haven't tested in rockbox) for pre, and another couple percent for post
17:55:12 Join pretty_function [0] (~sigBART@
18:02:42lorenzo92bertrik: fixed something more and simplified wm codec write!
18:05:43***Saving seen data "./dancer.seen"
18:13:24lorenzo92fm radio working fine even with the amplifier ;)
18:19:19lorenzo92bertrik: ouch! i've just found a bug, let me fix ^^
18:20:25 Quit DexterLB (Read error: Connection reset by peer)
18:24:54 Join DexterLB [0] (~dex@
18:43:50 Quit Strife89 (Ping timeout: 240 seconds)
18:50:34 Quit DexterLB (Read error: Connection reset by peer)
18:54:26 Quit ikeboy (Quit: ikeboy)
18:55:27 Join DexterLB [0] (~dex@
18:56:45 Nick SuperBrainAK is now known as DormantBrain (~andy@2001:470:8:a61::5f92:59a1)
19:03:48 Quit lorenzo92 (Ping timeout: 248 seconds)
19:25:47 Quit pretty_function (Remote host closed the connection)
19:26:17 Join pretty_function [0] (~sigBART@
19:30:49 Quit pretty_function (Ping timeout: 245 seconds)
19:42:16 Join lorenzo92 [0] (
19:49:17 Quit lorenzo92 (Remote host closed the connection)
20:05:45***Saving seen data "./dancer.seen"
20:27:18 Quit chrisjj (Ping timeout: 272 seconds)
20:45:52 Quit y4n (Quit: only amiga makes it possible)
20:52:13 Join tertu [0] (
20:56:45 Join lorenzo92 [0] (
20:57:37 Join ikeboy [0] (~ikeboy@
21:31:42 Join rela [0] (
21:31:47 Quit rela (Changing host)
21:31:47 Join rela [0] (~x@pdpc/supporter/active/rela)
21:36:05 Quit ikeboy (Quit: ikeboy)
21:40:45 Join Tukeke [0] (~Tukeke@unaffiliated/hlvsv)
21:44:44 Quit futurisk (Ping timeout: 246 seconds)
21:45:39 Join futurisk [0] (
21:48:33 Join Zagor [0] (
21:48:33 Quit Zagor (Changing host)
21:48:33 Join Zagor [242] (~bjst@rockbox/developer/Zagor)
22:01:12 Quit tertu (Ping timeout: 272 seconds)
22:05:46***Saving seen data "./dancer.seen"
22:06:20 Join tertu [0] (
22:20:08 Quit bluebrother (Disconnected by services)
22:20:13 Join bluebrother^ [0] (~dom@rockbox/developer/bluebrother)
22:24:06 Quit fs-bluebot (Ping timeout: 272 seconds)
22:26:00 Join fs-bluebot [0] (~fs-bluebo@
22:40:28 Quit lorenzo92 (Ping timeout: 272 seconds)
22:45:14 Quit rela (Read error: Connection reset by peer)
22:49:59 Quit amayer (Quit: Leaving)
22:58:38 Join lebellium [0] (
22:59:35 Join Cultist [0] (~CultOfThe@2601:d:9280:fc:8634:97ff:fe17:5dc3)
23:01:31 Quit tertu (Ping timeout: 252 seconds)
23:01:59 Join tertu [0] (
23:07:43 Quit kevku (Ping timeout: 256 seconds)
23:09:50 Quit Cultist (Ping timeout: 240 seconds)
23:23:26 Quit bluebrother^ (Ping timeout: 240 seconds)
23:23:44 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
23:30:14 Quit fs-bluebot (Ping timeout: 240 seconds)
23:33:51 Join Cultist [0] (~CultOfThe@2601:d:9280:fc:8634:97ff:fe17:5dc3)
23:34:29 Join fs-bluebot [0] (
23:39:26 Nick DormantBrain is now known as SuperBrainAK (~andy@2001:470:8:a61::5f92:59a1)
23:49:36 Quit Zagor (Quit: Leaving)
23:50:35 Nick SuperBrainAK is now known as DormantBrain (~andy@2001:470:8:a61::5f92:59a1)
23:57:04 Nick DormantBrain is now known as SuperBrainAK (~andy@2001:470:8:a61::5f92:59a1)

Previous day | Next day