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 2013-01-20

00:00:42saratogaif i highpass the signal and then upsample it so that its no longer critically sampled, I get something that looks like digital data
00:01:12saratogamodulated at pretty much 19khz
00:03:10pamaurywhich cutoff frequency do you use ?
00:05:27 Quit nxs (Ping timeout: 245 seconds)
00:07:05saratogai just did it by eye
00:07:10saratogaprobably about 15khz
00:07:46saratogaalso, if that data is correct, either the chip or the radio station has an absolute frequency error of only 0.9956
00:08:47 Join bertrik [0] (~quassel@rockbox/developer/bertrik)
00:09:09pamaurysaratoga: is that good ?
00:09:15saratogayeah its good
00:09:31saratogaprobably just means the clock is slightly faster then 38k
00:09:50saratogado you know matlab?
00:10:14pamauryno, I only know maple but I'm not sure it's quite as useful for this ^^
00:10:52pamauryand I used audacity to try a few things
00:11:16 Join Wardo [0] (
00:11:33pamaurySo I would need to implement a high-pass filter in software, right ? How do you do that ?
00:13:01saratogaoh wow i got it
00:13:08pamaurydo the discretization of an ideal high pass filter ?
00:13:15saratogawe have some in rockbox already
00:13:24saratogahold on, i'll put up a figure
00:15:15saratogared is the clock, blue is the signal
00:15:27saratogayou can see that it flips between in phase (1) and out of phase (0)
00:16:20pamauryamazing :)
00:16:25 Quit Wardo (Quit: Blarglarg)
00:17:08saratogahmm should probably check that those bits are at the right rate and not just random :)
00:19:03pamauryso they sample at 38kHz a 19kHz carrier, that's really the minimum you can do
00:19:11 Join TheSphinX^ [0] (
00:19:36 Quit TheSphin- (Read error: Operation timed out)
00:19:47saratogaits actually a 38100 Hz sampling rate
00:20:15saratogawell i get that a phase change happens every 64 samples, whereas the real rate should be 32 i think
00:20:21saratogaso maybe i'm missing something?
00:20:54saratoga38000Hz/1187.5bits/s = 32 samples per bit right?
00:21:37pamauryyeah, sounds correct
00:22:07saratogaoh wait that must have been two consecutive 1s
00:22:37saratogai found a transition that happens after about 32 samples
00:23:49pamauryis there a free version of matlab ?
00:24:49funmanGNU Octave ?
00:26:40pamauryok, I think I'll stop here for today
00:26:50pamaurysaratoga: many thanks for your help
00:27:26*gevaerts expects saratoga to tell us what radion station pamaury is listening to any moment now :)
00:29:54saratogawell i can demodulate it and all, but i have no idea what the bistream means
00:32:32pamaurywell differential encoder + NRZ + error correcting codes
00:33:07pamauryhey, the RDS spec actually has a nice diagram of how it works with pictures of the signals
00:35:11saratogaok i can spit out 1s and 0s
00:35:20saratogabut the SNR sucks, so i'm probably not scaling something right
00:36:31saratogaevery 32 samples, if the black trace is positive its a 1, negative a 0
00:37:21 Quit melmothX (Remote host closed the connection)
00:45:49saratogapamaury: do you want some data?
00:48:43pamauryyes :)
00:48:52 Quit Raptors (Ping timeout: 244 seconds)
00:49:27saratogano promises there aren't bit errors though as i think my code sucks :)
00:49:50saratogaactually i have no idea if any of that is correct
00:49:54pamauryrds is made to handle errors
00:51:03saratogaactually, that data is probably wrong
00:51:08pamauryanyway, the real time version of this will be much more tricky, I expect
00:51:18pamaurywhy ?
00:51:25saratogai'm just assuming that its exactly 32 samples per bit
00:51:31saratogabut theres probably some drift overtime
00:51:36pamauryhum, indeed
00:51:41saratogaso the first bytes are probably ok but i bet i drift over time
00:51:42*kugel also coded a (q)psk in matlab a while ago
00:52:04saratogayours is probably a lot more useful
00:52:18saratogai'm just mixing the carrier and signal, and checking to see if its one or zero
00:53:34kugelI did the same :p
00:53:52bertrikwhat does the rds spec say?
00:54:29bertrikI suppose it has a very description of the modulated signal and some hints on how to decode it
00:56:56pamaurybertrik: yes, "do it in hardware" ;)
00:58:22 Join Raptors [0] (
00:58:33saratogai think the problem with what i'm doing is that i just compute the carrier once for the whole signal, and then try to use it for a lot of points in time
00:58:53saratogaprobably this has to be done piece-wise, with the carrier estimated locally for each bit of incoming data
00:59:41saratogalooking at my data, it seems like at some points i have good SNR because the two clocks are well synced, and at other points they drift apart and my SNR degrades
01:00:09saratogaapparently phase is important to phase shift modulation
01:00:35kugelno shit!
01:01:15saratogalookning at the diagrams online, i guess thats why they bandpass out the incoming signal and mix it, that way you're always phase locked
01:04:17pamauryyou mean that they extract the carrier from the multiplexed signal ?
01:05:30 Quit enriched (Remote host closed the connection)
01:05:38pamauryhum, indeed that more or less what the official rds diagram suggests. In this case we cannot do that, we don't have the carrier. Stupid chip
01:05:51kugeldon't you usually have a pll at the receiver that locks into the phase of the inbound signal?
01:06:21kugel(of the carrier part, that is)
01:06:33saratogawe can get the carrier from the signal
01:06:34bertrikthe carrier should be exactly 3x the stereo pilot, with a specific phase IIRC
01:07:05 Join enriched [0] (~quassel@
01:07:22saratogaif you FFT the raw data, the strongest peak is the carrier
01:08:21saratogathere are two phases which represent the data, but they have the same frequency, and so we can just ignore that there is data by band pass filtering around the 19khz frequency
01:08:23 Quit Raptors (Ping timeout: 260 seconds)
01:10:55saratogaactually, does that work?
01:13:13saratogai wonder what happens if you get a bunch of zeros or ones in a row
01:14:40gevaertsThere's probably a system to avoid that
01:15:10 Join Raptors [0] (
01:15:13saratogait uses NRZ
01:15:17pamaurysaratoga: NRZ makes sure you have too many zeroes or one in a row, if it's the same thing as usb
01:15:23pamaury*don't have
01:15:42saratogai thought NRZ was the one where you can have an unlimited number in a row?
01:16:11saratogaoh i guess not
01:16:35pamaurybut can you really get the carrier from the data ? the rds data has very limited bandwidth. and you want to do that in real time anyway. Pfff, that's so easy to do correctly in hardware
01:16:56pamauryI don't understand why the chip was designed this way
01:17:40bertrikeach 16 bit word is encoded as a 26 bit symbol, for error detection/correction and I think for making sure that there are not too many 0's and 1's in a row as well
01:18:44 Quit bluebrother (Disconnected by services)
01:18:49 Join bluebrother^ [0] (~dom@rockbox/developer/bluebrother)
01:19:09pamaurysaratoga: what's the problem with too many 0 and 1 anyway ?
01:19:36gevaertsmiscounting them
01:20:57 Quit fs-bluebot (Ping timeout: 245 seconds)
01:22:13 Join fs-bluebot [0] (
01:32:38 Quit ender` (Quit: The three most dangerous things in the world are a programmer with a soldering iron, a hardware type with a program patch and a user with an idea.)
01:33:44 Quit bertrik (Remote host closed the connection)
01:49:39saratogaIIUC, the way this would work is that you'd have to bandpass filter the signal to recover the carrier, then compare the instantaneous signal to the carrier to see if its in or out of phase, but if theres a lot of one type or the other, the carrier will drift
01:50:23saratogamaybe theres a solution to that though?
01:53:45***Saving seen data "./dancer.seen"
01:58:08pamaurysaratoga: bandpass filter the rds signal ?
02:06:35 Join Katu1 [0] (
02:22:54pamaurysaratoga: the rds says: the subcarrier must be locked either in phase or in quadrature to the third harmonic of the pilot. What does locked in quadrature mean ?
02:39:06 Join orange_ [0] (43ddde44@gateway/web/freenode/ip.
02:40:08orange_Does anyone have an ipod classic that run off a cf card?
02:45:40orange_Is anyone using RockBox on an ipod classic with a compact flash card?
02:47:07[Saint]Such comments are unlikely to garner a response. The "right" way is "I am using <X> and experiencing problem <Y>".
02:57:14saratogapamaury: quadrature means 90 degrees
02:57:28saratogaso phase locked but with a 90 degree shift i guess
02:57:58 Part orange_
03:00:22saratogathe comments in the linux driver say that the 19k and 57k carriers aren't always locked
03:01:14 Join nateloaf [0] (
03:03:56pamaurywelcome to the real world :)
03:38:23 Quit Rower (Quit: Hmmm...)
03:46:08TheSevenrasher: if nobody else offers something, I might have CPU cycles to spare for rockbox builds
03:46:26TheSevenI'm currently not even running the normal build system though (didn't get around to setting it up)
03:46:48TheSevenoh, and no static IP over here. but 10mbit/s of upstream bandwidth at least
03:47:38rasherTheSeven: I'm fine with hosting the downloads as I've been so far, so bandwidth isn't much of an issue
03:48:44TheSevenany special requirements? I assume I need to have the rockbox toolchain and android SDK installed? NDK as well?
03:50:09[Saint]SDK/NDK, and for some reason I can't get compilation going right without having both openjdk6 *and* 7.
03:50:49rasherhm, I recall having to uninstall one of them
03:51:15rasherTheSeven: but yeah, mingw32 and openjdk - the sdk/ndk can just stay local to my user
03:51:48[Saint]the SDK, size wise.
03:52:00*[Saint] missed multiple words.
03:52:12[Saint]the SDK is bloody ridiculous now, size wise.
03:52:29TheSevendepends on which targets you install
03:52:42[Saint]No, no...just, by default.
03:53:01TheSeveni think the bulk of it is an AVD /system image for every installed android version
03:53:14[Saint]It extracts to ~300MB+ without installing any additional files.
03:53:35[Saint]NDK is near 1GB
03:53:46***Saving seen data "./dancer.seen"
03:53:49TheSevenoh, sure... sdk can be several gigs as well
03:54:07[Saint]for our purposes it stays around ~500MB
03:55:05TheSevenrasher: if you don't get better offers by tomorrow, contact me so that I set it up. the server would be an ubuntu server amd64 box with an ivy bridge i3 cpu and 8 gigs of RAM
03:55:51TheSevenoh, and tell me how much disk space you need so that i can set up lvm accordingly
03:56:09[Saint]TheSeven: dyndns to provide a fixed route?
03:56:18rasher[Saint]: not really necessary
03:56:18[Saint]re: no fixed IP.
03:56:31[Saint]No, but, trivial.
03:56:40TheSevenyes, it has a dyndns hostname
03:56:57[Saint]there we go, you do have "a fixed IP" :)
03:57:28TheSevenand it's running a little apache anyway, that we could reuse if wanted
03:58:11TheSevenactual ip behind that host name changes every 24h along with a few seconds of downtime
04:30:04 Join amiconn_ [0] (amiconn@rockbox/developer/amiconn)
04:30:04 Quit amiconn (Disconnected by services)
04:30:04 Quit pixelma (Disconnected by services)
04:30:04 Join pixelma_ [0] (pixelma@rockbox/staff/pixelma)
04:30:06 Nick pixelma_ is now known as pixelma (pixelma@rockbox/staff/pixelma)
04:30:06 Nick amiconn_ is now known as amiconn (amiconn@rockbox/developer/amiconn)
04:34:08 Quit pamaury (Ping timeout: 244 seconds)
04:36:23 Join TheSphinX_ [0] (
04:39:42 Quit TheSphinX^ (Ping timeout: 245 seconds)
05:01:32 Join PurlingNayuki [0] (
05:02:04PurlingNayukiI've got banned from the forum without doing anyting...
05:03:50saratogawhats your username?
05:04:06PurlingNayukiThat's yzflcyq
05:05:11PurlingNayukiI'd like to read a discussion in themes and open the forum only to got a banned message
05:06:01saratogaare you still banned?
05:06:34saratogaoh i see, you're on a host thats mostly spammers
05:06:59PurlingNayukiWhy's that happen.. is one of the largest spam hosts in the world, so i blocked it a while ago
05:07:39saratogaare you still banned?
05:08:02PurlingNayukiI've tried and yes, stil banned
05:08:46 Quit [Saint] (Remote host closed the connection)
05:09:39PurlingNayukiI don't think I can change my host...I'm now in mainland China and using ChinaTelecom as my ISP, behind the GFW
05:10:13saratogai've removed that ban, i didn't realize normal users were on it too
05:10:29saratoganot sure if you can post yet, might have to wait until morning in europe when one of the admins reads the logs
05:10:55PurlingNayukiThat's OK.
05:11:06PurlingNayukiThanks and I can just wait.
05:11:20saratogaperhaps one of the admins can check if anyone else of the several thousand people banned by that rule had any non-spam posts
05:11:53saratogaor just revert all of those bans
05:12:35PurlingNayukiYou know, when using ADSL users got different IP every time
05:12:49 Quit PurlingNayuki (Quit: CGI:IRC)
05:12:56 Join PurlingNayuki [0] (
05:14:13PurlingNayukiI own a site about Rockbox, chiefly for Chinese-speak users
05:15:58PurlingNayukiI can use the forum now and many thanks to you
05:16:35saratogaoh cool
05:16:54saratogai guess it removed all the bans when i disabled the rule
05:17:02saratogasorry again
05:17:41PurlingNayukiI searched for a keyword of '' and realize that it's the Reverse DNS for all the chinatelecom ISP users
05:18:17PurlingNayukiso I do advise you to check others.
05:18:45PurlingNayukiThere may be users banned but not knowing how to get obtain helps
05:32:56 Join TheSphinX^ [0] (
05:34:27 Quit TheSphinX_ (Ping timeout: 248 seconds)
05:40:39 Quit TheSeven (Disconnected by services)
05:40:47 Join [7] [0] (~quassel@rockbox/developer/TheSeven)
05:50:27 Quit tchan (Quit: WeeChat
05:53:50***Saving seen data "./dancer.seen"
05:59:19 Join [Saint] [0] (~saint@rockbox/user/saint)
06:00:27 Join TheSphinX_ [0] (
06:01:56 Join dfkt_ [0] (dfkt@unaffiliated/dfkt)
06:02:59 Quit dfkt (Ping timeout: 260 seconds)
06:03:11 Quit TheSphinX^ (Ping timeout: 264 seconds)
06:11:12 Quit PurlingNayuki (Quit: CGI:IRC (EOF))
06:12:37 Quit XavierGr (Ping timeout: 252 seconds)
06:15:34 Quit TheSphinX_ (Read error: Operation timed out)
06:22:58 Join TheSphinX^ [0] (
07:05:58 Quit Raptors (Ping timeout: 260 seconds)
07:10:30 Join TheSphinX_ [0] (
07:12:24 Join Raptors [0] (
07:13:29 Quit TheSphinX^ (Ping timeout: 252 seconds)
07:29:31 Join TheSphinX^ [0] (
07:32:13 Quit TheSphinX_ (Ping timeout: 260 seconds)
07:37:47 Quit SuperBrainAK (Quit: pbly going to sleep /_\)
07:44:10 Quit jhMikeS (Ping timeout: 256 seconds)
07:53:29 Join akaWolf [0] (~akaWolf@unaffiliated/akawolf)
07:53:51***Saving seen data "./dancer.seen"
08:15:07 Join eckoit [0] (~ryan@
08:45:51 Quit eckoit (Quit: eckoit)
08:50:28 Join kevku [0] (~kevku@2001:470:dcc6:0:69:69:69:69)
08:54:58 Join n1s [0] (
08:54:58 Quit n1s (Changing host)
08:54:58 Join n1s [0] (~n1s@rockbox/developer/n1s)
08:56:39 Join [Saint_] [0] (~saint@rockbox/user/saint)
08:57:37 Quit [Saint] (Ping timeout: 245 seconds)
09:04:53 Quit nosa-j (Ping timeout: 264 seconds)
09:10:23 Join nosa-j [0] (~m00k@
09:11:04 Nick [Saint_] is now known as [Saint] (~saint@rockbox/user/saint)
09:27:30 Quit n1s (Read error: Connection timed out)
09:51:29 Join melmothX [0] (~melmoth@unaffiliated/melmothx)
09:53:54***Saving seen data "./dancer.seen"
10:06:51 Quit zchs (Remote host closed the connection)
10:08:12 Join pretty_function [0] (~sigBART@
10:21:00 Join wodz [0] (
10:21:06wodzkugel: ping
10:21:43kugelwodz: ping
10:23:00wodzkugel: with your proposition of directly using fopen() and friends in plugins on hosted platform there is problem with paths
10:23:36kugeloh, right
10:24:29kugelnow that you say it, normal open() etc are also wrapped
10:24:43kugelthen discard that part of my comment
10:26:40wodzNow the final thing - do you have an idea how I could avoid separate header name (I mean that if someone uses stdio.h in plugin it would just work)
10:27:07wodzcurrently you would need #include "lib/stdio_compat.h" explicitly added.
10:28:23wodzkugel: ^
10:28:29kugelstdio probably conflicts with the core one, right?
10:28:41wodznot really
10:28:51wodzit conflicts with host's
10:29:33kugelin what way?
10:29:35wodzhmm, maybe include lib/stdio_compat.h from core stdio.h in case of plugins
10:30:22wodzhmm, wait I didn't try it after _ wrapping maybe it doesn't conflict this way
10:30:59wodzanyway apps/plugins/lib is not in path
10:32:54kugelindeed it isnt
10:33:49wodzcoming back to hosted platform - what do you think is better: 1) wrapping just fopen() (as we do for open() in hosted env) 2) proceed as I proposed on g# which means consistency between hosted and native but probably slower
10:34:11kugelI think an extra header file is okay (so porting effort is down from changing every call site just one line), but the underscores shouldnt be needed?
10:35:07wodzIf I wrap fopen() to fix path problems we could drop underscores and NOT compile wrapper for hosted targets
10:35:22kugelwith 1) we woud only wrap fopen() and the other function would be taken from the host?
10:35:47wodzyes this should work AFAIK
10:35:54kugelsounds better to me, yes
10:36:00wodzok, will try
10:38:11kugelfwiw, I thought if it might be useful in the core too (but with buffering)
10:38:31kugelbut it's harder to implement and probably not worth it
10:40:03wodzI don't think we need this in core.
10:40:20wodzNot that much external code is ported to the core
10:40:39kugelbut the buffering could boost performance in some areas
10:41:34wodzbut this wrappers do not provide buffering at all!
10:41:45kugelthat's why I said "(but with buffering)"
10:41:51wodzah ok
10:42:04wodzwe relay on buffering on fs level
10:43:32wodzis enough to precede the path with ROCKBOX_DIR in case of hosted?
10:43:38wodz*is it
10:45:48kugelfirmware/common/rbpaths.c has it for open()
10:45:56*wodz reading
10:47:11wodzhell, that is more complicated
10:47:58kugelcould export these the functions to plugins
10:49:28kugelif you call handle_special_dir() from the fopen() warpper it should just work
10:52:00wodzhandle_special_dir() is static
10:54:49kugelthat can be changed
10:56:59wodzlooking at it I am no more sure it is worth the hassle
10:57:22wodzthe current code on gerrit is quite simple after all
10:57:34kugelwhat hassle?
10:57:53kugelif you export this function then you need 15 lines of fopen(), and that's about it
11:00:12wodzexport like unstatic it and add to plugin api?
11:04:00wodzwell since we do not care much for plugins on hosted platforms is it really worth expanding plugin api? There is no other use for this function I can think of
11:04:52kugelit is, IMO
11:05:39kugelthe api is full of function that were added for a single plugin (and our policy is that adding a function is not a deal)
11:07:00wodzthats one of this polices which are not written down :-)
11:07:22kugelmost of our policies are implicit unfortunately
11:08:05wodzAnyway I see your point but I am not convinced thats the way to do.
11:09:24wodzwe could export this function only for hosted platforms
11:09:40wodzthat would make more sense
11:10:12kugelyes, sorry I didn't say that but I was thinking that since the beginning
11:10:21kugelsince it doesnt exist yet for non-hosted :)
11:12:42wodzwhat are the flags?
11:15:03kugeloh hmm, actually it exists only for APPLICATION
11:16:34 Quit Unhelpful (Ping timeout: 248 seconds)
11:16:41wodzit seems like NEED_WRITE is only needed in fopen() case right?
11:18:05kugelwhat do you mean?
11:18:19kugelit depends on the flags that are passed to fopen
11:18:47 Join Unhelpful [0] (~quassel@rockbox/developer/Unhelpful)
11:19:01wodzyes, but I don't need to specially check for file (IS_FILE) no?
11:19:37kugelcan you fopen() dirs?
11:20:07wodzprobably no
11:20:19kugelopen also sets IS_FILE unconditionally
11:21:10wodz"The fopen() function opens the file whose name is the string pointed to by path and associates a stream with it." So you cannot fopen dir
11:22:56 Join Rower [0] (
11:24:43wodzkugel: why is YPR0 en exception in many places?
11:25:37kugelit doesnt map <HOME> to the real home
11:25:53kugelfor the YPR0 the real home is known at compile time
11:30:52kugelif it's too much hassle for you now you can leave it alone, perhaps I can do the rest at a later time
11:31:27 Join n1s [0] (~n1s@rockbox/developer/n1s)
11:35:51 Quit n1s (Ping timeout: 252 seconds)
11:36:37 Join ender` [0] (
11:38:09 Quit ender` (Read error: Connection reset by peer)
11:38:20 Join ender` [0] (
11:50:51wodzkugel: rbpaths.c is not compiled for sim :/
11:53:58***Saving seen data "./dancer.seen"
11:54:03 Quit pretty_function (Remote host closed the connection)
12:07:58 Join pretty_function [0] (~sigBART@
12:30:59 Quit dfkt_ (Ping timeout: 248 seconds)
12:40:38JdGordon[Saint]: how does the .icons file work?
12:40:46JdGordonhow do you set the setting?
12:44:30 Join Wardo [0] (
12:55:48 Join y4n [0] (~y4n@unaffiliated/y4ndexx)
13:10:20kugelwodz: yes, I found that too
13:10:40kugelI forgot that sims io is wrapped in yet another place
13:17:30kugeluhh, I have a song that exposes heavy distortion under rockbox, but plays fine with vlc
13:17:41kugelon RaaA
13:19:31kugelandroid is fine, seems to be either an issue with my laptop or the C code of the codec (android uses asm optimzed code)
13:19:54 Quit pretty_function (Remote host closed the connection)
13:32:25 Quit ender` (Ping timeout: 240 seconds)
13:39:02 Join ender` [0] (
13:39:55 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
13:53:59***Saving seen data "./dancer.seen"
14:03:01 Quit wodz (Quit: Leaving)
14:29:28 Join n1s [0] (~n1s@rockbox/developer/n1s)
14:41:12 Join lorenzo92 [0] (
14:43:26lorenzo92kugel: remaining in the discussion of hosted platforms, I notice that under YPR0 memory isn't used enough, to express that better, I get a "no enough plugin memory" while trying to open some image files, also if there is a plenty of ram still to be used at any time (~12 mb if I correctly recall)
14:44:29lorenzo92moreover, HOME folder isn't mapped since we simply do a symlink at startup to /.rockbox -> /mnt/media0/.rockbox
14:48:20lorenzo92kugel: is this song affecting YPR0 too?
14:49:23 Join prof_wolfff [0] (
14:54:51pamaurysaratoga: can you post the code you used to process the rds data ?
14:55:38kugellorenzo92: I wouldnt think so (re: the song)
14:59:44lorenzo92kugel: and about plugin buffers? Can we somehow increase its size?
15:01:24kugelwhich plugin did you run?
15:01:49kugelah, imageviewer
15:02:15kugelshould be 512k (as for every target)
15:02:40gevaertsDoesn't imageviewer grab the audio buffer if it needs to?
15:03:37gevaertsgrep says it should
15:04:38lorenzo92kugel: uh oh! only 512k, no I understand...hum
15:04:43lorenzo92ah yeah can be
15:04:56lorenzo92in fact when a song is playing chances to get this error are higher
15:05:05gevaertsThat's weird
15:05:37gevaertsIf it doesn't have enough memory for an image it should stop playback and grab the audio buffer, possibly after asking
15:06:23lorenzo92but in any case, why no other ram is being used if the target has still much to give? hum I need to learn how buffers are allocated in rb ^^
15:07:09gevaertsHow much RAM does the device have, and how big is the audio buffer?
15:07:18kugellorenzo92: for RaaA there's a large static array (MEMORYSIZE mb big), other allocations are made from this array
15:07:53kugelincreasing MEMORYSIZE should work but keep in mind that the OS (and other processes, if any) also need some spare memory
15:11:58gevaertsAlso, I'd assume linux does some readahead, so there's a good chance the "wasted" memory will avoid some disk accesses anyway
15:13:52lorenzo92gevaerts: can be, but 12 mb are still free, well I have a look right now through uart
15:13:56lorenzo92just a second
15:14:17gevaertsActually, on RaaA I'd say we should look into redoing the audio buffer to mmap() the files instead, combined with careful use of madvise() :)
15:14:28*gevaerts waits for a volunteer to step up for that
15:15:07lorenzo92I will have time for this in february ;)
15:15:15lorenzo92while porting R1
15:15:54lorenzo92and indeed optimizations for RaaA are necessary, since there are lot of possibilities related to it
15:16:52kugelgevaerts: what would we gain?
15:17:20lorenzo92uh! stopping the process makes the alsa wrapper crash or something like that
15:17:24lorenzo92just to say ;)
15:17:58lorenzo92right now: MemTotal: 62416 kB
15:18:00lorenzo92MemFree: 12532 kB
15:18:09gevaertskugel: if things are properly tuned, optimal RAM use (the audio buffer wouldn't be fixed size any more, effectively). If things are not properly tuned, lower battery life, probably :)
15:18:47gevaertslorenzo92: is this just after boot, or during playback?
15:18:57lorenzo92during playback
15:19:11gevaertsDo you have swap?
15:19:19kugelthe ypr0 build has MEMORYSIZE=32 currently, iirc
15:19:24lorenzo92no swap
15:19:58lorenzo92killing rockbox, we have 39956kb free
15:20:11gevaertsWhat sort of numbers do you have for Buffers and Cached?
15:20:24lorenzo92let me do a pastebin
15:21:05gevaertsWithout swap you have to be careful, since OOM is rather easy to trigger
15:23:24gevaertsDo you have the blockdev tool? Playing with the filesystem readahead values might be a good way to increase RAM efficiency without increasing MEMORYSIZE (and therefore without increasing OOM risk)
15:24:23lorenzo92so it appears the allocation is done dynamically (correct) but is limited (not correct)
15:25:15kugelwell, it's a static array from rockbox perspective
15:25:27gevaertsI'd say the way rockbox works today, linux caching is not very useful
15:25:29kugelhow linux allocates it on process launch is a different story
15:26:31gevaertsSo I'd say increasing MEMORYSIZE to 48 or something like that might be worth a try
15:26:32lorenzo92it does copy-on-write I guess before executing
15:26:46lorenzo92gevaerts, yes I can
15:28:05gevaertsOf course you really need a battery bench for each setting you try :)
15:28:32*kugel doubts the effect is een measurable
15:29:00gevaertsI'd say what you want is the linux directory cache, but not the linux page cache, since we have our own system for that
15:29:54lorenzo92ok interesting
15:30:04lorenzo92so I should disable caching?
15:30:19kugelwell, i would leave a small page cache
15:30:24 Join TheSphinX_ [0] (
15:31:20gevaertsWell, not disable, just don't give it as much.
15:31:23lorenzo92kugel: indeed, just lowering it...let me see in /proc/vm
15:31:55gevaertsAnd without swap, one easy way to do that is grabbing more for rockbox :)
15:32:04 Quit TheSphinX_ (Read error: Operation timed out)
15:33:03 Quit TheSphinX^ (Ping timeout: 248 seconds)
15:34:30lorenzo92gevaerts: nice, thanks for the hint! dropping caches frees a plenty of ram
15:35:02lorenzo92i.e. echo 1 > /proc/sys/vm/drop_caches
15:35:47kugelit doesnt necessarily let us operate more effeciently though
15:36:04gevaertsWell, dropping caches at some interval won't
15:36:22lorenzo92no I need to set it indeed
15:36:25 Join TheSphinX_ [0] (
15:36:46gevaertsCurrently we're probably caching lots of stuff twice though, once in the audio buffer, and once (because it's been read recently) in the page cache
15:37:03gevaertsDuring normal use we're unlikely to read that data again
15:37:50 Join VanniX [0] (
15:37:50lorenzo92and in any case hw is fast enough :D
15:37:58gevaertsSo I'd say either we reduce the audio buffer and tell the kernel to aggressively read ahead (so we use the page cache as the main buffer), or we increase the audio buffer and thereby effectively tell the kernel not to cache so much
15:38:30lorenzo92I'd prefer the second, rockbox controls the device after all :D
15:38:38kugelI think the former is not very smart, unless we can give linux our playlist to readahead? :)
15:38:39gevaertsThe first needs research, the second is easy to try :)
15:39:05gevaertskugel: we can, if we want to. There's the readahead() system call
15:39:40gevaertsI'd go for a larger MEMORYSIZE first though
15:39:45kugelokay, but we have the code that does the same in place already
15:40:20lorenzo92I'm building a 48 mb build
15:41:12kugelif a 32mb build has 12mb free 48mb isn't going to work well, but worth a try
15:41:39lorenzo92kugel: it should, since freeing buffers seems freeing lot of memory
15:41:41gevaertskugel: that 12mb doesn't include the (imo useless) cache
15:41:58gevaertsSo those 12mb are *really* wasted and not even duplicates
15:42:47lorenzo92yes, exactly
15:42:59lorenzo92free memory
15:43:15kugelbt 48mb - 32mb is also more than 12mb :)
15:43:51lorenzo92yes I tought the same, but kernel should be smart enough or?
15:43:55gevaertsThe page cache also eats more than it should
15:43:55kugelthe paste shows linux has 40mb free before launching rockbox
15:44:29gevaertsYes, but it has lots of data in cache (and cache is *not* counted in those 40mb) that we likely won't ever need again
15:45:02kugelhow do you tell which cache to prioritize?
15:45:07gevaertsYou can't
15:45:08kugeltell linux*
15:45:27gevaertsYou can, but not without code
15:45:48gevaertsI'm not sure we really need any cache though
15:46:02kugeldirectory cache is nice to have
15:46:04gevaertsWhich other processes are running?
15:46:37*gevaerts decides to ask google for help
15:46:55VanniXgevaerts, in linux?
15:47:23VanniXwait, i can paste a ps
15:47:35gevaertsVanniX: on yp-r0
15:48:19lorenzo92as you see, nothing :D
15:48:28gevaertsOK, so not much :)
15:49:16gevaertsSo yes, what we ideally need to do is tell the kernel that the directory cache is important, the page cache is not
15:49:17VanniX <−− ps -l
15:50:47lorenzo92I'm trying the 48 mb build in a few
15:51:12VanniXMem: 23780K used, 38648K free, 0K shrd, 5392K buff, 11552K cached
15:53:48gevaertskugel: have a look at posix_fadvise(). We could (should?) call that for buffer filling with I think POSIX_FADV_SEQUENTIAL and POSIX_FADV_NOREUSE, and possibly POSIX_FADV_DONTNEED just before closing
15:54:00***Saving seen data "./dancer.seen"
15:54:23lorenzo92VanniX: you need to play a fikle
15:54:35gevaertsThat doesn't prioritise the directory cache as such, but with any luck it should help
15:54:35VanniXok :P
15:54:36lorenzo92otherwise buffer isn't counted ;)
15:55:17VanniXMem: 49808K used, 12620K free, 0K shrd, 120K buff, 11164K cached
15:56:59lorenzo92VanniX: the usb uart cable is great :D
15:57:12gevaertslorenzo92: do you have vmstat on the ypr0? If so, running "vmstat 1" (or some higher number if you like) and watching bi and bo should show io patterns which might correlate with disk spinups, which might avoid lengthy battery benches :)
15:57:52gevaertsOf course, running tools also impacts free memory, so you might want to adjust MEMORYSIZE a bit to compensate :)
15:59:24VanniXlorenzo92, now i'm with physical uart XD
15:59:29lorenzo92okay 48 mb is not good
15:59:53lorenzo92VanniX: I mean, physical through usb
16:00:47lorenzo92gevaerts: no vmstat
16:01:21lorenzo92I try with 40 mb
16:02:28lorenzo92okay one thing is clear
16:02:49lorenzo92without dropping cache we have ~43mb, after ~55!
16:03:02lorenzo92so lowering caching is a good point to do
16:04:22lorenzo92maybe I need to tune /proc/sys/vm/vfs_cache_pressure ?
16:04:48gevaertsThere are *many* things you can tune...
16:05:04lorenzo92yeah that's why I ask :D
16:05:12*gevaerts doesn't even want to guess which ones to try first :)
16:09:33kugellorenzo92: what happened?
16:09:55lorenzo92kugel: simply gets Killed
16:10:45kugelthat likely means not enogh memory to even start rockbox
16:11:15*kugel got confused a lot by this message during gsoc
16:13:24lorenzo92lorenzo92: also gets confused by virtual memory tunin'
16:15:29 Quit Wardo (Read error: Connection reset by peer)
16:16:27 Join TheSphinX^ [0] (
16:19:22 Quit TheSphinX_ (Ping timeout: 244 seconds)
16:20:11lorenzo92okay I need to go, I will go deep into this, very interesting! I want to get my 12 mb that are left alone!! XD
16:23:07 Quit lorenzo92 (Quit: ChatZilla 0.9.89 [Firefox 18.0/20130107224849])
16:26:13 Join esperegu [0] (
16:41:31 Join bertrik [0] (~quassel@
16:41:31 Quit bertrik (Changing host)
16:41:31 Join bertrik [0] (~quassel@rockbox/developer/bertrik)
16:43:34 Quit esperegu (Read error: Connection reset by peer)
16:43:52 Join ztf [0] (
16:44:06 Join esperegu [0] (
16:45:04ztfis clip v1 hanging a known issue?
16:45:28pamaurybertrik: I made some progress on the stfm1000 radio as you heard !
16:46:37pamauryI can successfully get sound. stereo/mono indicator doesn't seem to work or is unimplemented. And I'm looking at RDS.
16:47:00pamauryActually your work was amazing, I only needed to fix the dri code
16:47:35gevaertsztf: when does it hang? There are some known issues with USB for some people
16:47:42gevaertsAlso, which version?
16:48:01ztfclip v1
16:48:12*gevaerts assumes 3.12
16:48:13ztfit hangs during playback
16:48:33gevaertsThere was no 3.1.2, and if there had been, it would have been years ago :)
16:48:34ztfrequires a hard reset
16:48:51gevaertsI'd suggest checking the filesystem for errors
16:48:54bertrikpamaury: I took the existing code and tried to simplify it, but never got further than digital noise from it :)
16:49:31ztf3.12 oops
16:50:06pamauryI had to change the way dri was enabled. A comment in the linux code said that if the clock was enabled too late you would get noise. I also implemented a correct double buffering scheme and now it works. The code is still ugly but I'll fix that.
16:50:09ztfhow would one go about checking the filesystem?
16:50:32ztfi moved my music after installing rockbox btw
16:50:45gevaertsWhich OS do you use?
16:50:53ztflinux atm
16:51:38gevaertsOK, fsck.vfat /dev/sd<something>1
16:51:57kugelpamaury: what means dri in this context?
16:52:07pamaurydigital radio interface
16:52:13pamaurya propriatery saif if you prefer
16:52:29ztfaight ill give it a try
16:52:39*kugel doesn't know about either, but thanks
16:52:57pamauryit's a serial audio interface, so you can transport sound in digital form
16:53:40n1sthe radio outputs digital audio?
16:54:23pamauryand dri is a bit special in that it includes some synchronisation bits in the stream so that the SoC doesn't get lost, that's all basically
16:55:02pamauryand the streams is multiplexing four channels: L+R, L-R, RSSI and RDS
16:55:38ztfdosfsck 3.0.12, 29 Oct 2011, FAT32, LFN /dev/sdb1: 84 files, 16178/120752 clusters
16:55:52ztfdoesnt seem to tell much : /
16:56:38 Join zaphee [0] (
16:58:11bertrikpamaury: I think it wasn't entirely clear to me what the sample rate of the digital audio was, or how to configure it
16:58:12gevaerts84 files? That doesn't sound right
16:58:46ztfhmm now that i think about it..
16:58:52ztfi've got 99 tracks on it atm
16:59:33pamaurybertrik: you cannot really configure it, you can tweak the serial interface speed (because you can actually multiplex more or less than 4 channels) but in all cases the sounds is sampled at 44.1kS/s and rds is sampled at 38kS/s
16:59:56gevaerts.rockbox has around 250 files in it as well
17:00:30ztf: S
17:01:22 Join XavierGr [0] (~xavier@rockbox/staff/XavierGr)
17:02:06gevaertsYou could try running it with -v. Maybe that says something useful
17:02:16gevaertsPossibly even -l
17:03:33ztfoh boy, wrong device lol sorry
17:03:40ztfthats my sdhc card
17:03:50ztf /facepalm
17:03:51gevaertsThat would explain things :)
17:05:49ztf /dev/sdc: 538 files, 53613/62375 clusters
17:07:09ztfReclaimed 161 unused clusters (5275648 bytes).
17:08:09ztfFATs differ but appear to be intact. Use which FAT ? and theres this question
17:08:37gevaertsThat's basically just guessing. It's not possible to know which one is correct...
17:08:59ztfi tried both
17:09:20ztfperhaps i should try performing -r?
17:10:36gevaertsUnless you know FAT quite well, I suspect that's not really going to be very helpful
17:11:37gevaertsIf you have backups of everything, I'd actually recommend -a instead :)
17:12:15ztfaight, lets do it : P
17:13:50ztfPerforming changes.
17:13:51ztfoh well
17:13:59ztfthats not very informative
17:14:04ztfguess it didnt do anyting
17:14:28gevaertsYou could run it again. If it doesn't see the errors any more, it did something
17:14:54ztfthere was no Performing changes. this time
17:15:10ztfyeah here's hoping it did its magic
17:15:32gevaertsAnyway, the rockbox FAT code isn't as resilient as it should be against even slight filesystem corruption, so repairing the filesystem often fixes random issues
17:15:45ztfI see
17:16:25ztfwell, thanks for assistance
17:16:31ztf= )
17:16:59 Quit ztf (Quit: CGI:IRC)
17:22:49 Quit XavierGr (Disconnected by services)
17:22:51 Join XavierGr [0] (XavierGr@rockbox/staff/XavierGr)
17:30:15 Quit zaphee (Quit: Leaving.)
17:31:13 Join zaphee [0] (
17:32:14 Quit XavierGr (Disconnected by services)
17:32:15 Join XavierGr [0] (~xavier@rockbox/staff/XavierGr)
17:32:38 Join eckoit [0] (~ryan@
17:41:57 Join thomasjfox [0] (~thomasjfo@rockbox/developer/thomasjfox)
17:44:05 Part eckoit
17:50:33 Quit bertrik (Ping timeout: 260 seconds)
17:54:01***Saving seen data "./dancer.seen"
18:19:53 Join wodz [0] (
18:20:11wodzis there a way to force compilation of 32bit sim on 64bit system?
18:20:42funmanCC='gcc -m32' maybe?
18:20:56wodzyou mean in root Makefile?
18:21:08funmanfor example yeah
18:23:39wodzit doesn't link this way
18:23:54wodzobjects are 32bit though
18:24:30kugel-m32 should be sufficient, tried a make clean?
18:24:33funmanhm 'gcc' is used for linking instead of gcc -m32?
18:24:42funmanlibSDL 32bits is missing / not found?
18:26:12wodzit barfs on 1) tons of something like this: libuisimulator.a(buttonmaps.o)' is incompatible with i386:x86-64 output 2) undefined reference to `__divdi3' /__udivdi3
18:26:21wodzand yes I did make clean
18:26:57wodz-m32 is not passed to gcc when linking
18:31:20 Quit thomasjfox (Quit: Konversation terminated!)
18:34:01pamaurywodz: you need the 32-bit version of the libraries
18:34:20pamaurymainly pthread and sdl I think
18:35:47pamauryand I would put -m32 in GCCOPTS and LDOPTS
18:36:47wodzI leave it as installing 32bit of SDL is trying to wipe out half of my system
18:37:18kugelno multiarch?
18:39:56wodzwell I have quite a few 32bit programs installed and ubuntu should be smart enough so I don't know why it is so
18:40:25wodzanyway I am quite close to have functional forth kernel running as plugin :-)
18:40:26funmanwodz: you are on debian?
18:40:38wodzubuntu (xubuntu to be precise)
18:40:46funmando you have gcc-multilib?
18:41:28kugelwodz: how are you installing sdl? apt-get install libsdl1.2:i386?
18:41:53kugelthat should work since 11.10 or 12.04
18:42:37 Join AndChat|198900 [0] (
18:43:49AndChat|198900Pamaury: can you check that post on the zen in the new ports forum?
18:44:21AndChat|198900(Saratoga on phone )
18:46:54pamauryabout "Creative ZEN" ?
18:47:49AndChat|198900Iirc its a sigmatel very similar to the imx
18:48:16pamauryindeed, and the ZEN is very similar to the ZEN X-Fi which I own, iirc
18:49:10pamauryok, I'll post an answer as soon as this internet connection becomes again, I cannot even load the forum page
18:49:51wodzkugel: hmm, I tried to install libsdl1.2-dev:i386 - maybe thats the problem
18:50:04 Quit AndChat|198900 (Quit: Bye)
18:50:32kugelhm that should probably work as well
19:24:53 Quit the-kyle (Read error: Connection reset by peer)
19:25:00 Join the-kyle [0] (
19:26:22 Quit n1s (Quit: Ex-Chat)
19:28:58 Quit wodz (Quit: Leaving)
19:28:58 Join bertrik [0] (~quassel@rockbox/developer/bertrik)
19:54:04***Saving seen data "./dancer.seen"
20:12:21 Join TheSphinX_ [0] (
20:13:25 Quit TheSphinX^ (Ping timeout: 240 seconds)
20:14:22 Join TheSphinX^ [0] (
20:17:30 Quit TheSphinX_ (Ping timeout: 245 seconds)
20:18:52 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
20:23:50 Join SuperBrainAK [0] (
20:31:56 Join TheSphinX_ [0] (
20:34:08 Quit TheSphinX^ (Read error: Operation timed out)
20:40:56 Quit Raptors (Ping timeout: 245 seconds)
20:42:37 Quit TheSphinX_ (Ping timeout: 245 seconds)
20:43:01 Join Raptors [0] (
20:45:55 Join TheSphinX^ [0] (
20:48:17 Join lebellium_ [0] (
20:49:39 Join lebellium__ [0] (
20:50:09saratogasorry for matlab
20:50:27 Quit lebellium__ (Read error: Connection reset by peer)
20:50:55 Quit lebellium (Ping timeout: 248 seconds)
20:50:58 Join lebellium__ [0] (
20:51:12 Nick lebellium__ is now known as lebellium (
20:51:12saratogabasically, a really long (2^14 point) bandpass filter to pull out the carrier, then multiply the signal times that to see if its in phase or out of phase
20:51:25saratogaprobably need some more filtering in there to clean up noise
20:51:55 Join TheSphinX_ [0] (
20:52:45 Quit lebellium_ (Ping timeout: 252 seconds)
20:54:27 Quit TheSphinX^ (Ping timeout: 248 seconds)
21:09:18 Join Wardo [0] (
21:12:02 Quit esperegu (Ping timeout: 248 seconds)
21:21:14 Quit y4n (Quit: Do you like hurting other people?)
21:21:46 Quit TheSphinX_ (Ping timeout: 264 seconds)
21:27:06 Quit saratoga (Ping timeout: 245 seconds)
21:34:58 Join TheSphinX^ [0] (~briehl@
21:54:05***Saving seen data "./dancer.seen"
21:56:11 Join stripwax [0] (~Miranda@rockbox/developer/stripwax)
22:01:14 Quit akaWolf (Ping timeout: 256 seconds)
22:08:02 Quit SuperBrainAK (Ping timeout: 248 seconds)
22:16:30 Join TheSphinX_ [0] (
22:19:52 Quit TheSphinX^ (Ping timeout: 252 seconds)
22:25:54 Join Honkboy [0] (
22:27:38 Quit TheSphinX_ (Ping timeout: 260 seconds)
22:28:44 Quit Honkboy (Remote host closed the connection)
22:31:31 Join TheSphinX^ [0] (
22:34:09 Join Strife89 [0] (~Strife89@
22:34:12 Quit nateloaf (Quit: Leaving.)
22:34:31 Join TheSphinX_ [0] (
22:36:32 Join TheSphin- [0] (
22:37:15 Quit TheSphinX_ (Read error: Operation timed out)
22:37:35 Quit TheSphinX^ (Ping timeout: 248 seconds)
22:41:43 Join lorenzo92 [0] (
22:48:02 Join TheSphinX^ [0] (
22:51:16 Quit TheSphin- (Ping timeout: 276 seconds)
22:52:02 Join TheSphinX_ [0] (
22:55:00 Quit TheSphinX^ (Ping timeout: 252 seconds)
22:56:46 Quit stripwax (Quit: Miranda IM! Smaller, Faster, Easier.
23:07:16[Saint]JdGordon: Awesome work on the .icons file
23:07:33[Saint](and viewers.config, but that is a whole mess of broken)
23:11:05JdGordonthats a nasty file to work in! :)
23:11:12JdGordonfiletypes.c i mean
23:16:07pamaurywhy ?
23:16:25[Saint]It is, yes. I''m a little bit more savvy than I was the last time I looked there though.
23:16:36[Saint]I may not run out screaming this time.
23:21:16 Join SuperBrainAK [0] (
23:21:26 Quit melmothX (Remote host closed the connection)
23:24:13 Join Scromple [0] (~Simon@
23:24:46pamauryit doesn't look that bad imo
23:47:12 Quit zaphee (Quit: Leaving.)
23:54:07***Saving seen data "./dancer.seen"
23:55:39 Join eckoit [0] (~ryan@
23:55:45evilnick[Saint]: Any idea where I should put the contents of a for RaaAa?
23:56:10[Saint]I assume the device doesn't have a dedicated sdcard?
23:56:29[Saint]does /storage0 symlink to /sdcard or so? (it should)
23:56:50evilnickYes, and yes
23:57:24[Saint]Right, it goes in "/sdcard" then.
23:57:36[Saint](whereever that actually is, be it virtual or no)
23:57:59evilnickSweet, I think I've got it - is there a cabbiev2 as part of the .apk?
23:58:39[Saint]There is, but it should be ignored.
23:58:48[Saint]Oh, depends on the res actually.
23:58:52[Saint]there may well not be.

Previous day | Next day