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 2010-02-14

00:00:09saratogafor now I was just going to try and use the new fft with it
00:00:17saratogasince thats really easy and would get most of the speed up
00:00:21mtOh yes I remember now. I looked at that months ago and forgot about it. (bias)
00:01:19saratoga#define BIAS(x) ((x) + (bias*0))
00:02:35 Quit petur (Quit: Zzzz)
00:02:40waranMhm. I think I found the guy, but not sure ^.^
00:03:14saratogaemail him and ask
00:03:20waranThats what I did
00:04:40waranIf he turns out to be the original creator and allows me to steal his work I might be able to publish my personalized Brownstowne-Theme under CC-license on the rockbox themes pages. Would be awesome :)
00:05:09pixelmaI wonder what kind of answer that is. Such a post can easily be read as unfriendly though I give you that it could be overlooked. I just wish that people take themselves a little more time to answer and pay some attention
00:05:35saratogai'll try to be more careful in the future
00:06:53saratogaliba52 hasn't been updated since 2002
00:07:04saratogaanyone care if I kill dead code and such in it
00:08:39mtsaratoga: I think that bias is just there for floating point.
00:09:00saratogayeah looks like it
00:09:05saratogai took it out
00:09:24saratogai'll commit it to the mdct branch and then when we merge it'll get clean up
00:10:49saratogawhats the easiest way to run the sim remotely on a linux machine over xwindows and ssh? i know longer have direct access to an ubuntu machine
00:11:39mtAlso, this : #elif 1 (code that defines MUL(a,b).).... #else ... (Another definition of MUL(a,b)) .. What's that ??
00:12:53 Quit m3dlg (Ping timeout: 240 seconds)
00:12:56mtsaratoga: I used vnc when I needed to work remotely.
00:13:18 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
00:13:47freddybsaratoga: try ssh -X user@remote
00:14:26 Quit checker (Quit: CGI:IRC (EOF))
00:16:32saratogabut I'll still need an xwindows program for windows right?
00:18:23rasherBetter suited for #rockbox-community
00:18:36freddybdefinitly. I've used Xming before. (For this to work you have to have ssh_server (sshd) installed and running on remote. Which is not default in Ubuntu.)
00:19:32 Join Strife89 [0] (
00:20:06waranCurrently reading through the FAQ and I must say I really admire how much work was done to make rockbox possible. All this reverse engineering and stuff. Seems like rockbox is a real hackers masterpiece :D
00:22:26waran(by which I mean most of you guys, I suppose)
00:22:47Strife89waran: Most. :)
00:22:56Strife89waran: (Not me, I'm just a user. :) )
00:23:12 Quit tetsuo55 ()
00:23:45 Join stooo [0] (
00:24:53 Quit Olof- (Ping timeout: 240 seconds)
00:26:29saratogamrsomeone's todo list needs a "compare wav file" function in test_codec so we can more easily check if changes to the codecs introduced glitches
00:26:32saratogadoing it on pc takes too long
00:29:38 Quit punkt (Quit: Leaving)
00:29:49saratogasomeone deleted a locked thread with an explination of why I locked the thread
00:30:07saratogait would be nice if we didn't do that, since the whole point of putting the explanation is to remind people of the rules
00:33:23saratogastripwax, mt: do you care if I expose the fft in the codeclib.h header so I can use it in liba52?
00:33:35saratogafor now its a lot easier to use the fft directly then to figure out how to use the mdct itself
00:34:10mtI don't see a problem.
00:34:28saratogado you think its possible to put windowing code in the codeclib as well?
00:34:46saratogaliba52 should have an ASM optimized version, maybe sharing with another codec is possible
00:37:31saratogaliba52 also folds the windowing into the MDCT post rotation, which is probably a good idea on other codecs as well since it'll save a load/store per sample
00:37:34mtYou mean having separate windowing code/function in codeclib, that liba52 could use ?
00:38:25saratogaor maybe having an imdct_window function
00:38:37saratogathat computes the IMDCT and then windows the data during post rotations
00:38:45saratogausing a window function supplied by the codec
00:39:00mtYes I was just thinking of that too :)
00:39:02saratogai think most codecs could use that
00:40:28mtthe imdct function itself could be made aware of whether it should window or not, and what window to use. (instead of having 2 separate functions)
00:41:04amiconnsaratoga: "compare wav file" only makes sense for lossless
00:41:24TheSevenhaving separate functions might be marginallly faster
00:42:11saratogaamiconn: it would be useful for some changes to lossy as well as a sanity check
00:42:17 Join Schmogel [0] (
00:42:23 Quit Schmo (Ping timeout: 256 seconds)
00:42:30saratogabut yeah i suppose having it compute the RMS difference or something would be most useful
00:42:42mtTheSeven: Why ?
00:42:59 Join tetsuo55 [0] (
00:43:13 Quit tetsuo55 (Changing host)
00:43:13 Join tetsuo55 [0] (~eteu@unaffiliated/tetsuo55)
00:44:36TheSevenbecause not having to decide things will save some cycles
00:44:43*domonoky wants general support for subtracks. The way its done in nsf is nasty..
00:45:52*domonoky wanted to copy that for subtracks in the asap codec,but the i would loose the ability to seek :-/
00:47:13mtTheSeven: Missed that ! Thanks.
00:47:33saratogait'll just be one branch per call, typically about 200 calls a second, so not a big deal
00:53:37***Saving seen data "./dancer.seen"
00:54:17 Quit robin0800 (Ping timeout: 256 seconds)
01:00:13 Part froggyman
01:03:22 Quit utchybann (Read error: Operation timed out)
01:04:13saratogaheh liba52 doesn't even use IRAM for its imdct
01:04:45saratogawell the trig stuff is, but now the rest
01:08:23saratogaswitching over to the new FFT gives 21.94MHz decoding for the 192k sample
01:08:26saratoganot sure what it was before though
01:11:31 Join GeekShado_ [0] (
01:12:26 Quit GeekShadow (Read error: Connection reset by peer)
01:13:20mtsaratoga: Wasn't liba52's fft a bit faster ? or is the fft we're using now optimized ?
01:13:32saratogamt: the one we're using now has a lot of ASM
01:13:35saratogaso its probably faster
01:13:43saratogaand 22.76MHz
01:13:50mtoh ok.
01:13:51saratogaso we're 800kHz faster
01:14:00saratogabut i only switched long blocks over, not short
01:14:27saratogausing IRAM this thing could probably decode at << 20MHz on ARM
01:14:32saratogawe should really be using it for mpegplayer
01:18:01 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
01:19:13saratogaoh local variables are on the stack which is IRAM anyway
01:22:08 Quit moos (Read error: Connection reset by peer)
01:22:36 Join moos [0] (moos@rockbox/staff/moos)
01:25:35saratoga" warning: dereferencing type-punned pointer will break strict-aliasing rules"
01:26:08saratogai have no idea what that warning means, i'm basically just trying to cast from the liba52 complex type to the ffmpeg complex type which are functionally identical but delcared seperately
01:26:55 Quit piotrekm (Quit: piotrekm)
01:29:23 Quit dfkt (Quit: -= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.)
01:29:31saratogahmm easy enough to fix i guess
01:30:04 Join checker [0] (
01:30:19pamauryThis can happen when you "type-pun" a pointer, ie you access a variable with a pointer which is not of the same type. Typical with a union
01:31:02 Nick aaron_ is now known as Adnyxo (
01:31:25waranMerely bad style, but works ;)
01:31:41saratogathe structures are actually identical, they just have different member names
01:31:42checker%?bl<-1|0|1|2|...|N> using that conditional, |...| can be filled with as many sections "|-|" as desired?
01:31:47saratogaso i just renamed them
01:32:13pamauryI'm puzzled. Why isn't there a "title" object property with MTP...
01:32:37pamauryIsn't the title of an audio file one of its most important property oO
01:32:43waransaratoga, in way old pascal versions this would have been considered an error by the compiler :)
01:33:24 Quit bertrik (Quit: De groeten)
01:34:33saratogathis is c, a struct is just a funny type of array :)
01:35:02saratogaconverting short blocks makes no measureable difference in performance, probably because my sample doesn't really use them
01:35:08waranActually there is no array in c. Everything-is-a-pointer(TM) :D
01:35:42pamauryThat's false.
01:36:07waranthats why offset[pointer_var] works just like pointer_var[offset]
01:36:24pamauryIn some cases there is a fundamental difference between an array and a pointer. But these are subtle (and useful !)
01:38:19pamauryHum, that's should happen on community perhaps ?
01:38:56waranjust a tab away. Lets take it to community.
01:39:31saratogaok liba52 now uses the new codeclib in the mdct branch
01:40:14saratogaspeed up is less then 1 MHz, mostly because the ffmpeg and liba52 ffts are identical, the only difference being the ASM in the ffmpeg one
01:41:06saratogaliba52 is really interesting because its the only codec that i know of thats actually faster clock per clock on the gigabeat F then on PP
01:41:54saratogai'm assuming that means theres not very good IRAM usage in the codec, but looking at the imdct i'm not sure where the problem is
01:47:21saratogahuh looking at it more closely, it does look like its using IRAM extensively
01:47:28saratogai wonder what the gigabeat F does better
01:47:37saratogato overcome the IRAM advantage
01:48:31Unhelpfulwaran: it's not *always* only bad style, if two pointers of different types alias each other, but both have been loaded into registers, the compiler may assume that it can keep the value of *b in a register unchanged after you assign to *a.
01:49:16pamauryClearly strict aliasing is an optimiazation tool
01:50:23checkerwhat exactly is %bc
01:51:01*pamaury has object properties at least partially working for MTP using tagcache
01:51:12 Quit checker (Quit: CGI:IRC)
01:56:03 Quit perfectdrug_ (Quit: perfectdrug_)
02:03:18saratogai guess it doesn't make sense to convert too many formats to the new mdct until stripwax finishes with his init changes
02:03:25saratogait'll just be a lot of work that'll have to be undone later
02:08:18pamauryIs there any way to get the sample rate of a file using tagcache tags or not ? iirc, it's related to the frequency in metadata but in tagcache, I'm can't find an equivalent
02:08:56saratogapamaury: you could call the metadata parser on it I suppose
02:09:28pamauryYes of course but it's slow so if I can avoid it it's better
02:09:38saratogaalthough i'm curious why libmtp needs to know what the sample rate of files already on the disk is?
02:09:47saratogaerr mpt
02:10:09pamaurylibmtp doesn't it but I'm trying to implement the properties that are usually implemented in MTP
02:10:15pamaury*doesn't need it
02:10:22pamauryAnd sample rate is one of them
02:10:26saratogai thought it just checked the files as they came in and then didn't need that once they're on the disk
02:13:40mthmm .. avoiding the reversal of the spectral coefficients in the imlt in atrac3 only saves 0.11 MHz :/
02:15:23saratogai remember looking at requantization in atrac last summer and thinking we could speed that up a bit
02:15:52saratogathe tonal component coding thing happens too rarely to make much difference in performance, but the rest of it could be speed up
02:16:10 Quit waran (Quit: Miranda IM! Smaller, Faster, Easier.
02:16:16saratogaand of course theres the little issue with it not being real time on coldfire
02:17:50mtIs it still not realtime ? I thought Buschel's optimizations last summer fixed that ?
02:19:02saratogai don't think he did anything for coldfire
02:19:07saratogaIIRC he only has an arm target
02:20:37mtfound it :;revision=22561 .. The amazing +639% !
02:21:39 Quit anewuser (Quit: =ooo ϢINTER ϾHIP 5iVE is OOON!!)
02:23:16tetsuo55makes me wonder if that patch was ever upstreamed to ffmpeg
02:23:32saratogathats why i was thinking it wasn't real time
02:23:39saratogaperhaps some files are slower then others
02:24:37saratogatetsuo55: it won't be, its not really applicable to ffmpeg
02:25:06tetsuo55its a lossy speedup?
02:25:31saratogano its fixed point only
02:25:48mtah I see ..
02:26:04saratogaor at least not obviously applicable to ffmpeg
02:27:11tetsuo55doesnt look that obvious to me
02:27:19mtMaybe that user was using an old revision . the thread was in october, ~2 months after the optimizations.
02:27:47tetsuo55if it increases performance for any random case, without decreasing accuracy or perfoemance elsewhere, i'm sure it will be accepted upstream
02:29:02 Join domonoky1 [0] (
02:29:35 Quit domonoky (Ping timeout: 265 seconds)
02:29:56saratogatetsuo55: it optimizes fixed point math
02:30:01saratogaffmpeg doesn't even do fixed point math
02:30:23tetsuo55it doesnt?
02:30:25saratogaso its not really applicable to them
02:31:16 Quit domonoky1 (Read error: Connection reset by peer)
02:32:11mtShould I commit that small optimization to mdctexp and trunk, or just trunk, so that mdctexp would be just for the mdct stuff ?
02:32:29saratogai was wondering that
02:32:39 Join perfectdrug [0] (
02:32:40saratogais it in a file we'll need to change to use the new mdct?
02:33:04saratogathe ARM ltd profiler looks really nice
02:33:15saratogai wonder if theres a way we can get them to donate a copy or two
02:33:17tetsuo55saratoga > i just checked the ffmpeg site through google, seems like there is a big interest in fixed point math, the devs try to have all codecs have 2 codepaths, fixed and floating
02:33:37tetsuo55theres even a fixed point math port from rockbox to ffmpeg on the to-do list
02:34:05tetsuo55but not for this particular codecd
02:34:29mtit's in atrac3.c .. it's now using the new mdct in the mdctexp branch, so we'd just have to change those few lines when merging back.
02:35:27tetsuo55would make it easier to stay in sync with ffmpeg trunk too (assuming it gets accepted)
02:35:32saratogatetsuo55: do you have a link about the fixed point stuff?
02:35:53saratogaheh 2006
02:36:04tetsuo55ignore the dates
02:36:35tetsuo55lots more here:
02:37:14saratogawell when they get around to adding a fixed point decoder we should see about giving them optimizations for it
02:37:23tetsuo55anything that improves ffmpeg is a win in my book
02:37:33tetsuo55saratoga dont expect that
02:37:57tetsuo55ffmpeg has a clear policy, that they dont really work on feature requests, and that stuff like this has to be submitted as a patch
02:38:12tetsuo55rockbox clearly has a fixed point local version, that could be upstreamed
02:38:12saratogayes I know, i've spent a good deal of time over there
02:38:38tetsuo55and its probably not just atrac, but many other codecs
02:39:04tetsuo55may sound obvious, but ideally all codec updates will occur in the ffmpeg trunk, then get synced into rockbox
02:39:46tetsuo55i have a similar problem with my project, unfortunately my patches are really only usefull in my project, as they are api hooks
02:40:08mttetsuo55: They have a fixed-point cook patch that's been sitting in there (interesting patches) for ages, and we actually used that patch in rockbox - it decodes 200+% realtime btw - and it has yet to be committed.
02:40:09tetsuo55but if we find anything like memory leaks, etc.. we quickly submit a patch upstream
02:40:22tetsuo55i want that
02:40:22mts/in there/in their/
02:40:58mtWell, You have it in rockbox already. :)
02:41:02tetsuo55i assume you harras tem all the time in #ffmpeg-dev
02:41:04tetsuo55i dont
02:41:11*tetsuo55 has a D2
02:41:29tetsuo55i only play with rockbox in the sandbox on windows
02:42:04tetsuo55i use cook audio ALL the time on the pc though
02:42:19tetsuo55and that codec needs a lot of TLC
02:42:31LloreanWouldn't trying to keep all of our code in sync with theirs by submitting it to them and updating off of their version just create a ridiculous bottleneck in the rate at which codec improvements could occur unless all Rockbox codec contributors had ffmpeg commit access too?
02:42:56tetsuo55only 1 rockbox dev really needs commit access
02:43:19LloreanThis 1 dev being magically available 24/7/365?
02:43:29tetsuo55but yes, the bottleneck means you need a local ffmpeg branch that you can keep working on, while you wait for the offical ffmpeg trunk to catch up
02:43:50tetsuo55i have to go to bed now, more here
02:44:08LloreanAlso doesn't that mean more work for Rockbox devs to fit the code into ffmpeg, rather than ffmpeg devs who know their code doing that work?
02:44:32tetsuo55ideally things would go smoothly on irc
02:44:32 Quit pamaury (Quit: abort();)
02:44:43tetsuo55and snippets of code would turn into commits swiftly on their side
02:45:05tetsuo55instead of slowly in a mailing list that nobody cares about
02:45:27saratogai don't think trying to convert all of ffmpeg to fixed point makes much sense
02:45:35saratogait'd just make a mess of their codecs
02:45:43tetsuo55i agree, just the parts your already doing anyway
02:45:47saratogacombined fixed/float codecs are ugly and difficult to work with
02:46:03tetsuo55i would guess rockbox has a large pile of patches, that get applied to ffmpeg in some way
02:46:14saratogawe don't actually use ffmpeg itself
02:46:32saratogawe do adapt [by rewriting] some of their code though
02:46:33LloreanI'm just not really sure I see the 'benefit' of creating additional work to integrate code to another project that has its own devs perfectly capable of working with us to do that.
02:47:26tetsuo55saratoga so you only use small bits of ffmpeg, in your own codebase?
02:47:50saratogaa few of our codecs started out as ffmpeg codec, and one ffmpeg codec started out as a rockbox codec, but we do very different things
02:48:04tetsuo55Llorean > the idea is that you change something, to the default ffmpeg file, and then provide that patch upstream, and the ffmpeg devs take it from there, and ask you for input when needed
02:48:29tetsuo55Llorean > i do agree that its hard to swallow, to have to do the whole thing yourself
02:48:39Lloreantetsuo55: That also doesn't really reflect the real-world situation
02:48:50tetsuo55unfortunately it doesnt :(
02:48:52LloreanAs saratoga said, we're not really a port of ffmpeg. They're not really, realistically, "upstream"
02:49:04saratogatetsuo55: what is involved in converting a codec to fixed point?
02:49:05tetsuo55yeah, i just understood that just now
02:49:22LloreanIt's a good idea to try to learn about the situation before you make suggestions on how to 'improve' it. ;)
02:49:42tetsuo55i wrongly assumed it was similar to what my project does, include ffmpeg and wrap around that (with some minor patches to the ffmpeg source)
02:50:09tetsuo55saratoga > LOTS of beer
02:50:45saratogaif thats your answer then maybe you shouldn't be trying to tell people who have some idea what they're talking about how to do things
02:50:48tetsuo55Llorean > i based it off the few svn logs :P
02:51:29 Quit toffe82 (Ping timeout: 240 seconds)
02:53:11tetsuo55saratoga > i dont claim to have any idea how to do that, my mistake was that i misinterpretid the things i saw as a "local copy" where in fact it seems to be a "fork", and then only for a handfull of codecs
02:53:11 Quit saratoga (Quit: CGI:IRC)
02:53:17 Join saratoga [0] (
02:53:35tetsuo55[02:53] <tetsuo55> saratoga > i dont claim to have any idea how to do that, my mistake was that i misinterpretid the things i saw as a "local copy" where in fact it seems to be a "fork", and then only for a handfull of codecs
02:53:39***Saving seen data "./dancer.seen"
02:54:09 Join toffe82 [0] (
02:54:14saratogatetsuo55: if you have no idea what you're looking at its still not a good idea to give advice
02:55:11saratogaalso, if people are using lots of words you don't understand, thats another clue
02:55:18saratogahope that helps
02:55:39tetsuo55maybe, but the generic idea still stands, it would be nice, if this stuff was available upstream in ffmpeg(with lots of support from that side)
02:56:22 Join Rob2222 [0] (
02:56:37LloreanA lot of what we've got doesn't even make sense as a patch against their tree.
02:56:52LloreanIt really makes more sense to find someone who's familiar with their code and knows where fixed point codecs would fit within it.
02:56:53 Join CaptainKewl [0] (
02:58:01saratogahence my original comment that most of it isn't applicable to ffmpeg
02:58:46tetsuo55didn't most codecs used to be fixed point anyway?
02:58:52tetsuo55some time ago
02:59:09tetsuo55and now transitioned into all being floating
02:59:59LloreanThe transition from mostly fixed to mostly floating point code happened before compressed audio codecs like MP3 really took off
03:00:11 Quit Rob2223 (Ping timeout: 272 seconds)
03:02:08tetsuo55maybe im misunderstanding, but google is giving me fixed point decoders for almost all popular formats
03:03:01saratogaprobably ours
03:04:49tetsuo55saratoga > that might be the cause yeah
03:05:00tetsuo55lots of closed source stuff too
03:05:34tetsuo55but i can clearly see the rockbox patches being ignored in those search results too
03:06:15tetsuo55so its not for lack of trying from this side!
03:06:37 Quit toffe82 (Ping timeout: 258 seconds)
03:07:30 Join toffe82 [0] (
03:07:33saratogaall the results on that page for open source decoders are in rockbox as far as I can tell
03:08:15tetsuo55makes sense
03:09:19tetsuo55thanks for your serious responces
03:11:06*tetsuo55 sleeps
03:18:11 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
03:20:42 Nick fxb__ is now known as fxb (
03:21:01 Quit TheSphinX^ (Ping timeout: 246 seconds)
03:21:34 Join toffe82_ [0] (
03:22:02 Quit efyx_ (Remote host closed the connection)
03:22:43 Quit toffe82 (Ping timeout: 258 seconds)
03:25:47saratogawow the mailing list hates my email formatting
03:26:11 Join toffe82 [0] (~chatzilla@
03:29:12 Quit toffe82_ (Ping timeout: 264 seconds)
03:30:08 Quit saratoga (Quit: CGI:IRC (EOF))
03:30:14 Quit akur (Quit: Leaving.)
03:40:36 Join TheSphinX^ [0] (
03:42:22 Join Nimbus [0] (
03:43:26 Join toffe82_ [0] (
03:43:42 Quit Schmogel (Ping timeout: 256 seconds)
03:46:52 Quit toffe82 (Ping timeout: 240 seconds)
03:47:16Nimbusis anyone in here?
03:50:21mtNimbus: If you want to ask something please do, if someone is able to answer you they will.
03:57:27 Quit Nimbus (Quit: CGI:IRC (Ping timeout))
03:58:22 Join Nimbus [0] (
04:03:52 Quit Nimbus (Quit: CGI:IRC (Ping timeout))
04:09:21 Join teru [0] (
04:16:01 Join FlynDice [0] (
04:17:29 Quit TheSeven (Disconnected by services)
04:17:43 Join The_Seven [0] (~theseven@rockbox/developer/TheSeven)
04:17:55 Nick The_Seven is now known as TheSeven (~theseven@rockbox/developer/TheSeven)
04:18:14 Quit Tomis (Quit: Tomis)
04:22:53 Quit Sajber^1 (Read error: Connection reset by peer)
04:38:04 Join Tomis [0] (~Tomis@
04:39:41 Join Barahir_ [0] (
04:43:21 Quit Barahir (Ping timeout: 256 seconds)
04:46:41 Quit amiconn (Disconnected by services)
04:46:43 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:47:05 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:47:45 Quit pixelma (Disconnected by services)
04:47:46 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:48:04 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:51:09 Quit Adnyxo (Ping timeout: 256 seconds)
04:52:24 Quit stooo (Ping timeout: 260 seconds)
04:52:57 Nick fxb is now known as fxb__ (
04:53:40***Saving seen data "./dancer.seen"
05:04:29 Quit mt (Quit: ChatZilla 0.9.86 [Firefox 3.5.7/20091221164558])
05:06:57 Join stooo [0] (
05:11:23 Quit GeekShado_ (Quit: The cake is a lie !)
05:23:04 Quit Strife89 (Quit: Bed.)
05:26:52 Join kramer3d_ [0] (~kramer@unaffiliated/kramer3d)
05:29:13 Quit kramer3d (Ping timeout: 264 seconds)
05:32:43 Join S_a_i_n_t [0] (S_a_i_n_t@
05:33:00 Quit S_a_i_n_t (Client Quit)
05:33:18 Join S_a_i_n_t [0] (S_a_i_n_t@
05:42:57 Join Riku [0] (
05:42:58 Quit Lss (Read error: Connection reset by peer)
05:43:27 Nick Riku is now known as Guest93909 (
06:15:38*S_a_i_n_t is poised for emplanation
06:15:53S_a_i_n_texplanation rather
06:31:15 Part toffe82_
06:42:01 Nick kramer3d_ is now known as kramer3d (~kramer@unaffiliated/kramer3d)
06:44:18tmztLlorean: mobile computing wasn't huge then though, arm drove back to soft approximations over fp
06:45:16 Join kramer3d_ [0] (~kramer@unaffiliated/kramer3d)
06:45:16tmztI would personally love to have fixed point codecs in ffmpeg, gstreamer, etc. for my embedded projects on slow fp arm chips
06:45:23tmztembedded linux
06:46:10tmztffmpeg already has multiplr codec variants selectable at runtime
06:47:12 Quit kramer3d (Ping timeout: 264 seconds)
06:53:44***Saving seen data "./dancer.seen"
06:56:18JdGordonfreddyb: seet, all targets build?
07:02:00*S_a_i_n_t wonders how checking out a 'clean' svn revision reports being 24***:24***M
07:02:56freddybI did actually build them going through it. But there's a lot of targets and my automagic C checker is broken.
07:06:19S_a_i_n_tI wonder what I broke...svn is having a hissy fit at me.
07:07:19freddybJdGordon: is there an automatic way to build all the targets?
07:07:41JdGordonthere is a script somewhere to do it
07:10:28S_a_i_n_tcan it mess things up somehow if the connection drops while checking out a souce revision?
07:11:11S_a_i_n_tit's all I can think of, I deleted the old folder, checkout out a new revision...but it's al messed up.
07:13:02 Nick kramer3d_ is now known as kramer3d (~kramer@unaffiliated/kramer3d)
07:14:29freddybJdGordon: I'll put Samsung models in and then run The only problem I've had building stable targets was if I had a typo in the button_map instance. The last conditional is a catch all emtpy set.
07:16:25 Quit CaptainKewl (Ping timeout: 245 seconds)
07:16:34LloreanS_a_i_n_t: Are you using real linux, cygwin, or vmware, or something else?
07:16:49JdGordonfreddyb: ok, let me know when you are done and I'll commit it
07:17:38S_a_i_n_tcygwin...(yeah, yeah :P)
07:18:33S_a_i_n_tdeleting the old source folder should be sufficient no matter how bad I mess the tree
07:19:02LloreanCygwin can pose line ending problems. Are you using the console SVN command or something like TortoiseSVN?
07:20:00freddybJdGordon: Will do.
07:20:12S_a_i_n_texample of weirdness: checking out a brand new revision, and it stopping dead on " svn: Failed to add file 'manual/rockbox_interface/main.tex': an unversioned file of the same name already exists"
07:20:30S_a_i_n_twhich definately didn;t alredy exist...
07:26:08JdGordonhere we go.....
07:27:25*S_a_i_n_t is killed by suspense...
07:27:58JdGordonTransmitting file data ....................
07:28:10JdGordonis svn on the fritz
07:28:55S_a_i_n_tthat may explain my recent 'piss-offs' with trying to check-out...
07:33:46JdGordonok wierd, the checkin shows up on the main page.. I wonder whats broken
07:36:45 Quit kaniini (Quit: E.T. go home.)
07:44:03JdGordonoh bloody hell... game over for the ondio sp's rombox
07:44:22 Join kaniini [0] (
07:45:07 Quit Horscht (Ping timeout: 256 seconds)
07:45:49 Join Horscht [0] (~Horscht2@xbmc/user/horscht)
08:02:00S_a_i_n_thasn't it been a huge effort to keep the ondio's rombox alive fr as long as you have?
08:07:13 Quit phanboy4 (Ping timeout: 256 seconds)
08:22:05 Join robin0800 [0] (
08:25:16 Quit panni_ (Quit: ( :: NoNameScript 3.81 :: ))
08:37:04 Quit robin0800 (Remote host closed the connection)
08:48:35 Quit kramer3d (Ping timeout: 258 seconds)
08:51:59 Quit freddyb (Remote host closed the connection)
08:53:47***Saving seen data "./dancer.seen"
09:29:54 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
09:38:58 Join Kitr88 [0] (
09:41:52 Join bertrik [0] (
09:42:41 Quit Kitar|st (Ping timeout: 240 seconds)
09:43:13 Quit Kitr88 (Ping timeout: 246 seconds)
09:44:59 Join ender` [0] (
09:49:30 Join Kitar|st [0] (
10:05:26 Join flydutch [0] (
10:11:23 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
10:13:38 Quit kramer3d (Quit: Leaving)
10:38:15 Quit dude187 (Read error: Operation timed out)
10:53:51***Saving seen data "./dancer.seen"
10:58:16 Join utchybann [0] (
11:01:50 Join xavieran [0] (
11:05:20 Join petur [0] (
11:05:20 Quit petur (Changing host)
11:05:20 Join petur [0] (~peter@rockbox/developer/petur)
11:22:39 Join TopyMobile_ [0] (
11:26:11 Quit TopyMobile__ (Ping timeout: 240 seconds)
11:39:45 Join DerPapst [0] (
11:47:44 Join stripwax [0] (
12:11:06 Join stoffel [0] (
12:13:52 Join Buschel [0] (
12:14:26Buschelanyone with a coldfire target willing to test a atrac codec optimization? patch ->
12:14:43BuschelI need a short run with svn against the patched version
12:14:51Buschelmaybe pixelma? ;o)
12:20:38 Join Sajber^ [0] (
12:21:38gevaertsBuschel: what exactly do you need tested? Also, is there a test file somewhere I can use?
12:24:31Buschelgevaerts: I need a speed test via the test_codec plugin
12:24:51BuschelI am a bit puzzld there are no *.rm test files on
12:25:17gevaertsno free encoder maybe?
12:30:53Buschelsolved ;)
12:31:36 Join Farthen [0] (
12:31:41 Join robin0800 [0] (
12:35:49pixelmamt once linked me to a few (from some mplayer page or so?). Can't remember exactly unfortunately :\
12:36:22gevaertstesting on plain svn now (I assume you want to have both)
12:36:46Buschelpixelma: same for me. I have the files, but no link...
12:36:52Buschelgevaerts: yep
12:36:55 Quit Adub- (Read error: Connection reset by peer)
12:37:19 Join Adub- [0] (
12:38:31gevaertsDo you want me to listen to them as well, to see if the codec still works properly?
12:39:44Buschelyes, to be safe. butyou will only need to listen very briefly. if something is going wrong you ears would bleed ;)
12:40:30 Join Schmogel [0] (
12:41:23Buscheltoo many typos today... :/
12:41:45gevaertsThey seem to play OK
12:41:54 Join pamaury [0] (
12:43:26Buschelthanks! +8-10%, but still not fully realtime...
12:44:02Buschelbut there is still some room left for tweaking :)
12:44:11peturprogress is always good
12:46:02Buschelgevaerts: what target did you use?
12:46:18peturcertainly not h120 :)
12:47:08stripwaxBuschel - atrac not realtime on coldfire?
12:48:05Buschelok. does the codec stil compile for you when changing line 60 in atrac.c (from IBSS_ATTR_LARGE_IRAM to IBSS_ATTR)?
12:48:33stripwaxhave you tried the mdctexp mdct ?
12:48:42Buschelstripwax: it is realtime on coldfire targets with large iram.
12:49:12stripwaxah, which coldfire targets have 'small' iram? i have an h120
12:49:19Buschelstripwax: no ,it did not try mdctexp, but i have read it will save another 1.5 MHz on arm. so, let's check it for coldfire as well :o)
12:49:43stripwaxBuschel - I think mt reported an 11% speedup, or something (can't remember if that was for atrac though)
12:50:01stripwax(that wasn't on coldfire though)
12:50:03Buschelstripwax: the "large" one is CPU == MCF5250
12:50:19stripwaxBuschel - ok so the h120 has a 'small' one I guess
12:50:26gevaertsBuschel: no
12:50:31gevaertsIRAM full
12:50:44Buschelgevaerts: too bad
12:51:09*gevaerts goes out to buy more IRAM
12:51:26Buschelhehe :)
12:51:39stripwaxsaratoga - (for the logs) - I don't understand why liba52 can't just use codeclib mdct? Why is only the fft bit applicable?
12:52:08Buschelthe easiest way to optimize coldfire targets is to move stuff to iram
12:53:53***Saving seen data "./dancer.seen"
12:56:17 Quit S_a_i_n_t (Quit: [St.] has exited mIRC)
12:56:35 Join S_a_i_n_t [0] (S_a_i_n_t@
13:00:11 Quit Farthen (Quit: ChatZilla 0.9.86 [Firefox 3.6/20100115144158])
13:05:49 Quit stripwax (Quit:
13:10:09 Join mt [0] (~mtee@rockbox/developer/mt)
13:15:23 Join perfectdrug_ [0] (
13:18:41 Quit perfectdrug (Ping timeout: 240 seconds)
13:26:41 Quit shaggy-h (Ping timeout: 240 seconds)
13:27:23 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
13:33:46Buschelgevaerts: does this patch compile for your h300? ->
13:34:57Buschel:o) is it faster?
13:36:07*gevaerts wonders where he put this usb cable
13:36:54mtgevaerts: You sent it to me, remember ?
13:38:02gevaertsDid I? I thought I sent it to funman!
13:39:21pamauryHum, tagcache_fill_tags does not fill the length entry of mp3entry. I think I'll fix that in a second
13:41:55Buschelgevaerts: hmm, same as with first patch
13:47:52pamaurysvn is really slooooooooooooooooooow today
13:49:45 Quit bmbl (Ping timeout: 245 seconds)
13:51:05moospamaury: Slt, how goes your dircache rework, is it finished? I didn't test yet (not dev env here those last weeks)
13:51:38pamaurymoos: hello
13:52:15moospamaury: waw it tested enough?
13:52:16pamaurymoos: I have two pendings orthogonal dircache rework: 1) Remove this awful 30kb buffer in dircache (I have a patch in FS) 2) Remove any use of a buffer of size MAX_PATH on the stack. I have some code but not finished.
13:52:45 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
13:53:01pamaurymoos: you mean the one I commited more than one week ago ?
13:53:32moosoh ok, still working on it... If you need testing on big discs (aka big cache) I could
13:54:15moospamaury: the complete scan rework. I didn't find time yet to fix my laptop problems (ie no dev env)
13:55:12pamauryAh. I tested it and got not change. JdGordon also tried it and apparently it is working. I can't seen any reason for this to fail except if my trick with the FAT code is wrong but if think it's ok.
13:55:30pamaurymoos: but yes, if you can try it on a big disc this would be great
13:56:21moosdo you get improvements? I mean: is the scan faster now?
13:57:02moospamaury: I will retry to get e dev env here, sinon, you will have to build me a beast build patched
13:57:22pamauryOn my device I can't see a difference because I haven't got some many files. That's why a test on a big disc would be useful.
13:57:44pamauryI'm not sure it will scan faster with this because beasically it does the same thing but uses less memory
13:58:31moosdid you get Slasheri's attention about it?
13:58:54pamauryThe only possible improvement is the old code did depth first scan, this one does breadth first search so if the directory is contiguous, perhaps this is faster but it's unlikely.
13:59:09pamauryI haven't seen Slasheri for more than one week
13:59:43moospamaury: ok, I will say you how things became here...
14:00:41pamauryDid someone notice that JdGordon multifont patch broke Ondio SP on the build table ? (region FLASH is full (/home/robert/rb/builds/build-archosondiosp/rombox.elf section .rodata) )
14:01:15moosyup he advised that the rombox is now out
14:02:49moos"game over" for it to use his words :)
14:03:01moos*tu use
14:03:28*moos hides from amiconn's ammo :)
14:03:43gevaertsI wonder if it wouldn't be possible to regain some space by e.g. loading bitmaps from disk after boot
14:03:46S_a_i_n_tyes, I believe it was something along the lines of "bugger, its game over for the ondio sp's rombox..."
14:04:34pamauryI guess ondio sp doesn't use dircache...
14:05:17gevaertsit doesn't
14:06:06pamauryHum. Maybe I'll go through a little review of the code to see if I can gain some place. Does it use tagcache or not ?
14:07:05gevaertsapparently, yes
14:09:01Buschelgevaerts: next try -> does it compile and is it any faster?
14:10:41Unhelpfulcould just make multifont a bigmem-only feature :P
14:12:05gevaertsdoes it even actually work on lowmem? Isn't the skin buffer smaller there as well?
14:13:53 Quit teru (Quit: Quit)
14:13:59gevaertshm, on ondio, the bitmaps are only 940 bytes, so loading those after boot isn't goint to achieve much
14:14:01 Join teru [0] (
14:14:11S_a_i_n_tare commits not showing in channel for some reason?
14:14:30moosno CIA bot anymore
14:15:14gevaertsNo real change
14:16:06Buschelgevaerts: yes, a pitty. so, I better leave the iram free for more efficient future options.
14:16:08 Join efyx_ [0] (
14:25:20gevaertspamaury: tagcache uses 27K on the archoses, so yes, there might be room there
14:25:45gevaerts*27K binsize
14:27:18tetsuo55moos > CIA is down
14:29:10pamaurygevaerts: from what I saw, tagcache uses an enormous stack (16Kb !!!) so there is some work to be done in this area perhaps.
14:29:48gevaertspamaury: improving that will no doubt be good, but that won't fix rombox unfortunately
14:29:55 Quit petur (Quit: real life)
14:30:35pamauryTrue. But there perhaps are other areas to optimize. tagcache is kind of a fragile masterpiece so it's not the easiest one to tweak.
14:32:05gevaertsOne real long-term solution for rombox could be the ability to load parts of the binary from disk. That's probably not going to be very practical though
14:32:29 Join jeffp [0] (
14:34:02jeffphey guys, i was just wondering whether anyone's working on rb for sansa fuze v2
14:34:02amiconngevaerts: Loading bitmaps after boot kinda defeats the purpose of the splash screen
14:34:20 Join Farthen [0] (
14:34:39amiconn*especially* when running from flash
14:34:54gevaertsamiconn: I was more thinking about the usb logo and the icons, but anyway I think that this sort of thing is not worth it for only a few hundred bytes
14:35:39pamaurygevaerts: I know a place where there is memory: in dir_uncached using the same FAT trick, one can divide the size of DIR_UNCACHED by two at first sight (if I'm correct)
14:37:50 Join dfkt [0] (dfkt@unaffiliated/dfkt)
14:38:45Unhelpfulsurely reducing stack, buffers etc won't help code fit in flash?
14:38:45 Quit robin0800 (Read error: Connection reset by peer)
14:39:47pamauryDon't know but I'm listing areas where there is some things to take (in a general way)
14:40:14amiconnUnhelpful: Of course not
14:40:24pamauryWhat uses flash memory ?
14:40:46amiconnIn fact the binsize increase for multifont is relatively moderate, given what it adds (unlike some other features...)
14:41:13Unhelpfulpamaury: code. read-only data. not much else, i'd expect.
14:41:17amiconnpamaury: Code, read-only data and initial values for data
14:41:40 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
14:43:21gevaertsamiconn: do plugins change between rombox and rambox?
14:43:42*gevaerts suspects not
14:43:51amiconnAhem, no
14:44:31amiconnRombox, standard in-flash rombox and disk-based rockbox (.ajz/.mod) all come in the same package...
14:45:43 Quit Farthen (Ping timeout: 256 seconds)
14:45:57gevaertsWe could consider changing rombox builds to be separate builds, without e.g. tagcache. That wouldn't be a very nice solution though
14:48:56amiconnAnd it wouldn't work
14:49:24amiconnIf you leave out tagcache (or any other non-trivial feature), language and voice ids change
14:50:33gevaertsyes, they'd have to be fully separate builds. Not good
14:50:35amiconnThat means loading disk-based rockbox (which is required before flashing if rombox is currently running) would mess up the whole ui
14:50:59gevaertsYou could have an equivalent disk-based rockbox with it
14:51:47gevaertsThen people would just have to choose when installing: full rockbox with all features, or rockbox light with less features, but longer runtime
14:52:05linuxstbMove the "optional" feature strings to the end?
14:52:19gevaertsthat would still break the plugin API
14:52:47*gevaerts doesn't like this sort of "solution" at all
14:53:03amiconnRuntime isn't really different on te Ondios
14:53:18pixelmathere are still two targets that have rombox currently and for others it has been broken for quite some time (different time depending on the target)
14:53:23amiconnBut the larger buffer has other advantages, e.g. being able to view larger jpegs etc
14:53:41pixelmaor using a better quality voice
14:53:54***Saving seen data "./dancer.seen"
14:54:20amiconnIn fact runtime might be *slightly* worse with Ondio rombox
14:55:16amiconnThis is because the RAM has a 16 bit data bus, while the ROM only has an 8 bit bus. Hence code execution from ROM is slower, and the cpu gets less sleep
14:55:29 Quit Guest93909 (Ping timeout: 240 seconds)
14:55:58amiconnOn the hdd-based archoses, the savings from having less disk spinups more than make up for this
14:57:14amiconnOf course, code running from IRAM isn't affected, and is faster than from RAM. IRAM is 32 bit, but there isn't much of it...
15:00:57Buschelgevaerts: last try ->
15:03:05 Quit moos (Ping timeout: 276 seconds)
15:04:10 Join moos [0] (moos@rockbox/staff/moos)
15:05:43 Quit tetsuo55 ()
15:07:19 Join robin0800 [0] (
15:10:37gevaertsBuschel: no change
15:10:47 Quit robin0800 (Client Quit)
15:11:15 Join robin0800 [0] (
15:14:17 Nick fxb__ is now known as fxb (
15:14:31gevaertsBuschel: I have to go now. If you need more tests, you'll have to look for another victim^Wvolunteer
15:15:02Buschelgevaerts: thank you for your time!
15:16:35Buschelbtw, found the link to free atrac samples ->
15:25:11 Join tetsuo55 [0] (
15:25:29 Quit tetsuo55 (Changing host)
15:25:29 Join tetsuo55 [0] (~eteu@unaffiliated/tetsuo55)
15:30:01 Join akur [0] (
15:32:47 Join Lss [0] (
15:36:05moosre pamaury, tested your dircache patch. Working as plain svn (RAS).
15:37:05pamauryAnd what about dircache thread usage ?
15:37:22moosall is working as excepted, same scan time, same buffer size
15:37:54 Join phanboy4 [0] (
15:37:54*moos forgot to to check with svn (I will)
15:38:35pamauryNormally this patch should increase dircache stack usage so I would like to now if t's safe or not :)
15:39:19moosif you are curious, the cache here take 20s to process, and my buffer is around 1,47mo, for 22200+ elements
15:39:43moos120gb disc and gigabeast btw
15:39:49mooshehe, I will compare
15:39:56pamauryok thanks a lot
15:40:50pamauryIs there a way to know from how much rombox overflow the flash area on ondio SP ?
15:40:57moosno problems, thanks to you to works on :) (or to be brave enough to work on... to pe more precise ;)
15:42:58moosbtw I noticed that the "scaning disc" splash was in english (I do remenber to saw it too last time I tested your commited changes).
15:43:21moosI will check if is not french specific (i.e my mistake :)
15:45:27 Join Adnyxo [0] (
15:45:57pamauryamiconn, Unhelpful : does any of you know a way to know how much binsize we would have to free to have rombox fit the flash ? (becuse the error message does not give any information)
15:48:56Buschelto finally reach realtime for atrac on coldfire mr.someone should add some asm for the iqmf_dewindowing
15:49:56amiconnmoos: The localisation of boot splashes got broken with one of the theming commits. Another thing JdGordon or kugel need to fix :\
15:50:19 Quit S_a_i_n_t (Ping timeout: 256 seconds)
15:50:34 Join Scary [0] (~AreaFileU@
15:50:42moosamiconn: duh :/ (I was checking with another translation)
15:50:44amiconnIt's not just the dircahce scanning splash, but also db commit, out of extension array etc
15:50:54*moos is annoyed by the boot time also :(
15:51:18moosJdGordon don't trust use and wants stats compare :(
15:51:39amiconnBuschel: Doesn't atrac have similar complexity as mp3?
15:52:33Buschelamiconn: no, atrac uses very computing intensive qmf-filterbanks
15:52:54amiconnHmm, ok
15:52:59Buschelamiconn: at least 48 muls per sample (!)
15:53:13*amiconn wonders whether he could do something there
15:53:18bertrikjobec, can you try another battery bench on your clip after explicitly enabling and disabling the radio?
15:53:20amiconnFilters... and asm :)
15:53:23Buschelamiconn: and there is an imdct as well
15:54:13TheSevenpamaury: I guess it's some 500 bytes, but no, you can't see it directly.
15:54:16 Join S_a_i_n_t [0] (S_a_i_n_t@
15:54:33Buschelamiconn: yes, the dewindowing function is quite dumb, should be easy to implement in asm −− if one is able to write cf-asm :o)
15:54:35TheSevenIIRC newer LD versions tell you how much more memory they would need
15:55:57pamauryTheSeven: I have a weird way to found but I don't like it but if there isn't a clean way to do it, I have no choice
15:58:08moospamaury: going back to plain svn, and no functionnal changes as far I can tell, except of course the stack usage change (on the debug to 1%−−>72%). Your code changes seems safe at least :)
15:59:38pamaurymoos: the 1% is probably here because dircache loaded the cache from the disk. Can you try with plain svn and rebuild cache ? (this 1% is impossible otherwise)
16:00:27moospamaury: amiconn is maybe one of the only one user/who knows about rombox around here
16:01:10 Quit teru (Quit: Quit)
16:01:26pamauryHum, rombox would need 548 free bytes to fit the flash
16:02:40 Quit Sajber^ (Read error: Connection reset by peer)
16:03:10 Join moos_ [0] (
16:04:01pixelmadepends on which rombox you are talking about (which target)...
16:04:38pamauryondio SP
16:04:49 Quit moos (Ping timeout: 252 seconds)
16:04:50 Nick moos_ is now known as moos (
16:05:09 Quit moos (Changing host)
16:05:09 Join moos [0] (moos@rockbox/staff/moos)
16:05:11pixelmaI know, just saying that this is not the only target that would be interesting
16:05:37 Join GeekShado_ [0] (
16:05:56pamauryIs it the only one which doesn't fit in flash ?
16:07:53 Quit GeekShadow (Ping timeout: 240 seconds)
16:09:16 Quit moos (Ping timeout: 246 seconds)
16:09:18 Join moos_ [0] (
16:09:22 Nick moos_ is now known as moos (
16:09:35pixelmarombox doesn't fit on any other bitmap Archos anymore (just the compile errors have been silenced) and making it fit on the others is more far off. The reason is that others have more features that are possible (e.g. the OndioFM has a radio and recording, the Recorders also have backlight). I'm impressed you want to help there anyways though :)
16:09:43 Quit moos (Changing host)
16:09:43 Join moos [0] (moos@rockbox/staff/moos)
16:11:13pamauryPerhaps there is nothing do to but at least I can try.
16:12:24 Join moos_ [0] (
16:12:36pixelmathe worst are probably the FM-Recorders/ RecorderV2s
16:13:07moos_pamaury: comment added to the task.
16:13:16pamaurymoos_: I just saw it ;)
16:13:45pixelmaanother difference between Ondios and all Recorders is the RTC
16:14:33 Quit moos (Ping timeout: 256 seconds)
16:14:35 Nick moos_ is now known as moos (
16:14:47 Quit moos (Changing host)
16:14:47 Join moos [0] (moos@rockbox/staff/moos)
16:16:18 Join CaptainKewl [0] (
16:18:00 Join moos_ [0] (
16:19:05 Quit moos (Ping timeout: 256 seconds)
16:19:10 Nick moos_ is now known as moos (
16:19:17 Quit moos (Changing host)
16:19:17 Join moos [0] (moos@rockbox/staff/moos)
16:22:47 Join moos_ [0] (
16:23:48 Quit moos (Ping timeout: 264 seconds)
16:23:58 Nick moos_ is now known as moos (
16:24:15 Quit moos (Changing host)
16:24:15 Join moos [0] (moos@rockbox/staff/moos)
16:29:12 Quit moos (Ping timeout: 264 seconds)
16:31:57 Quit FlynDice (Remote host closed the connection)
16:32:14 Join FlynDice [0] (
16:34:00 Join moos [0] (moos@rockbox/staff/moos)
16:36:13 Join Jungti1234 [0] (
16:38:16Jungti1234long long time no see
16:38:18domonokywow, svn is slow today: last commit: 4kbytes transfered in 3m,10s :-)
16:39:01pamauryyeah, same for me
16:41:45 Quit n17ikh (Ping timeout: 256 seconds)
16:44:39 Quit Jungti1234 (Quit: CGI:IRC (EOF))
16:46:35 Join n17ikh [0] (
16:46:41pamauryhehe, I just won 16 bytes on ondio SP ... :)
16:48:17rasherHave people forgotten about the manual for WPS tags?
16:48:33 Quit phanboy4 (Ping timeout: 256 seconds)
16:51:05danopia_domonoky, git ftw ;)
16:51:45domonokydanopia_: git doesnt help if the repository i want to commit to is svn :-)
16:52:04danopia_i can mirror rockbox onto git :P
16:53:06domonokydoes some one know what happens if i load a cfg with settings which only exist for color targets on a non-color target ?
16:53:55domonokyare the non-existing settings safely ignored ? (thinking about checking theme cfgs on the theme page)
16:53:56rasherI also seriously doubt the slowness there was scm related
16:53:58***Saving seen data "./dancer.seen"
16:55:00rasherMore likely a holdup with the cia notification
16:55:08rasherSince cia appears to be gone
16:55:29domonokyalso what happens with non existing setting values ? for example "selector type: bar (color)" on a greyscale target ?
16:56:15domonokytrue, a timeout in the commit hooks is probably the cause for this slowness.
16:59:53AlexPcia is down
17:00:05AlexPAll the bots have disappeared and the site is down
17:00:37AlexPAnd Rockbox has a bit mirror already, but it is checkout only (obviously)
17:00:52 Quit Adub- (Read error: Connection reset by peer)
17:01:03 Join Adub- [0] (
17:01:47danopia_yea CIA is down
17:02:40danopia_i'd offer my replacement (FBI) but i'd probably have to add hacks for it to work with your repo
17:03:01danopia_i can't recieve email withotu some advanced mailserver goodnes but i think i can do XML-RPC
17:04:28jobecbertrik: i could try to remember to run one this night
17:06:07 Quit Scary (Ping timeout: 256 seconds)
17:17:38 Join panni_ [0] (
17:23:24 Join Loggerin [0] (~tanjango@
17:24:53 Quit Loggerin (Client Quit)
17:30:28 Quit fxb (Read error: Connection reset by peer)
17:35:49 Quit akur (Quit: Leaving.)
17:39:42 Quit moos (Quit: ChatZilla 0.9.86 [Firefox 3.6/20100115144158])
17:41:32 Quit bmbl (Quit: Bye!)
17:47:54 Join fxb [0] (
17:50:00 Nick fxb is now known as fxb__ (
17:50:02 Nick fxb__ is now known as fxb (
17:54:26amiconnpixelma, pamaury: Actually there is still one target where rombox is available: the Archos Player
17:55:10amiconnAnd the warnings/errors haven't been silenced, the rombox build step has been deactivated for the targets where it doesn't fit anymore
17:55:16pixelmawhich is no bitmap Archos... and there are also flashed H100s (which I meant earlier with "two")
17:55:52*amiconn overlooked the word 'bitmap'
17:56:06Unhelpfuldomonoky: you can commit locally on git. ;)
17:56:27amiconnThe H1x0 has no space problem at all (1MB flash instead of 256KB)
17:56:57amiconnAlthough swcodec rockbox is considerably larger than hwcodec (and the larger and 2bpp bitmaps add to that)
17:57:18domonokyUnhelpful: thats not what i want todo :-)
17:58:15Unhelpfuldomonoky: you want to be able to commit, i told you how you can commit. what's the problem? ;)
17:59:05 Quit martian67 (Ping timeout: 258 seconds)
17:59:10domonokyUnhelpful: no i want to make my changes live, the commit is only the tool for it. :-)
17:59:58Unhelpfulyou can do that too! mirror your git repo somewhere public. ;)
18:02:18domonoky"somewhere" is the wrong place.
18:05:24 Quit tha (K-Lined)
18:06:04pamauryIs rombox compressed in any way ?
18:07:21pamaurywow, it that valid C: [0 ... REC_NUM_FORMATS-1] = AFMT_UNKNOWN ???
18:08:13 Join Jaykay [0] (
18:08:27 Join JdGordon1 [0] (
18:08:47Unhelpfulpamaury: it can't be, it's executed directly from flash.
18:09:01pamauryah true ;) I'm so stupid
18:09:08 Join Guest57645 [0] (
18:09:56Unhelpfulthat might be another thing you could try re: initdata, though... rather than looking for individual initdata arrays that could be calculated, just compress the whole lot. :)
18:10:34 Join petur [0] (
18:10:34 Quit petur (Changing host)
18:10:34 Join petur [0] (~peter@rockbox/developer/petur)
18:11:08pamauryUnhelpful: you mean, compress all initdata and at runtime, uncompress them to RAM ?
18:11:33Buschelwhat is happening if the script returns weirs version like "r23773:24648M-100214"? where does the first "r23773:" come from?
18:12:45Unhelpfulpamaury: why not? initdata is used to provide the initial values of writable arrays, isn't it? so it needs to be copied to RAM anyway, doesn't it? it's at least worth seeing if compressed initdata + a decompressor will fit.
18:13:03 Join tomers [0] (
18:13:03Unhelpfulassuming it's not compressed already, of course...
18:14:52pamauryThat works only for initdata which is RW no ? Because otherwise will put it in .rodata and that will never end up in ram
18:15:14amiconnThere isn't that much int data
18:15:34amiconnAlthough we could also compress all .icode, .irodata and .idata
18:15:53amiconnNot that much either - SH7034 has 4KB of IRAM
18:16:35Unhelpfulamiconn: is "part of the data is compressed" perhaps something that could be done by the bootloader? i gather that bootbox is responsible for decompressing non-rombox firmwares... or am i mistaken?
18:16:56amiconnNot bootbox, but the flash loader
18:17:20amiconnThe flash boot on archos is a bit more modular than on the other targets
18:18:37pamauryWhat exactly does the boot time looks like on an archos ? First there is the flash boot and then ?
18:19:42 Join moos [0] (moos@rockbox/staff/moos)
18:20:56amiconnIt depends on whether the archos is flashed (flashing is optional, and not all boxes are flashable), and where you have a box with or without bootrom (the latter variant is rare)
18:21:53pamauryif it's flashed and with a box ?
18:22:39Unhelpfulmaybe we should work further on making features that make no sense on archos optional... the bitmap loader already doesn't support scaling on mono targets, but we could probably make a smaller bitmap loader specialized for loading bitmaps to mono.
18:23:14JdGordonwhat's the point? we are just going to be back here in a few weeks anyway
18:23:19 Join robin0800_ [0] (
18:23:33JdGordonit broke just before 3.5 branch
18:24:02amiconnThe standard way (without rockbox in flash): (1) CPU boots from its internal rom (it's a mask rom, not changeable). (2) The boot rom starts the archos firmware in flash. The loader that's part of it copies the rest of the firmware to RAM, then jumps there. (3) The firmware looks for an on-disk update (archos.mod for Players, ajbrec.ajz for the others). If it exists, it loads and starts that one
18:25:09 Quit robin0800 (Ping timeout: 258 seconds)
18:25:11 Quit JdGordon1 (Ping timeout: 252 seconds)
18:25:22amiconnOn a flashable box, we replace the in-flash firmware with 3 components: our flash loader, and two firmware images, which can bei either compressed or uncompressed
18:27:22 Join MethoS- [0] (~clemens@
18:27:46amiconnBoot sequence changes as follows: (1) CPU boots from its internal rom (it's a mask rom, not changeable). (2) The boot rom starts our flash loader. (3) The flash loader checks buttons, and does one of three actions depending on that. (a) start the second firmware image. This is the default, full rockbox, which is updated by rockbox_flash.rock
18:29:35amiconn(b) start the first firmware image. This is the backup - bootbox, a cut-down rockbox that just handles charging and usb. It is there in case the default image is broken. (c) start minimon, a mini serial monitor for debugging/testing
18:33:52pamauryI think I will go back to my dircache optimizations, that's easier ;)
18:38:00 Join stripwax [0] (
18:40:03amiconnEach of the two firmware images can bei either compressed (ucl) or uncompressed. In the latter case it is execute inplace
18:40:09amiconnBootbox is compressed, of course
18:41:09 Join sudoman [0] (
18:50:38sudomanrbuitilqt 1.2.5 is quitting unexpectedly when i try to open it. 1.2.4 works.
18:50:46sudomanthis might help:
18:51:06sudomanos 10.5.8
18:51:11sudoman* os x
18:51:13 Quit S_a_i_n_t (Ping timeout: 258 seconds)
18:53:08 Join S_a_i_n_t_ [0] (S_a_i_n_t@
18:53:58domonokysudoman: arg, not again. We had the same error with 1.2.4 on osx 10.4 :-/
18:54:00***Saving seen data "./dancer.seen"
18:55:39domonokybluebroth3r: ping ?
18:57:14 Join phanboy4 [0] (
18:59:46 Join Strife89 [0] (
19:03:17sudomanthe buttons used to force a ipod 3G reboot and then to enter disk mode are different than explained on
19:03:27 Join shai [0] (
19:03:51Jaykaysudoman: it's a wiki, you can change it :)
19:04:02sudomanhaha. thanks
19:04:27 Quit HBK (Read error: Connection reset by peer)
19:04:30 Quit phanboy4 (Read error: Connection reset by peer)
19:04:58 Join phanboy4 [0] (
19:05:32 Join HBK [0] (
19:05:43 Join Lss__ [0] (
19:06:04 Join hd [0] (
19:06:04 Quit hd (Changing host)
19:06:04 Join hd [0] (~jd@Wikipedia/HellDragon)
19:06:13 Quit jd (Read error: Connection reset by peer)
19:06:21pixelmasudoman: as far as I know it's correct (menu + select to reboot and then select + play)
19:06:42sudomanfor me, neither of those combos worked
19:06:59sudomani did play and menu then back and next
19:07:16sudomanmenu and select may work in the rb firmware, but i'm not sure
19:07:17 Quit lyngaas (Ping timeout: 272 seconds)
19:07:21 Quit shai (Read error: Connection reset by peer)
19:07:31 Join shai [0] (
19:07:37 Quit Lss (Read error: Connection reset by peer)
19:08:47 Quit Adub- (Read error: Connection reset by peer)
19:08:58 Join Adub- [0] (
19:09:48pixelmaah, 3rd gen (and older) might be different
19:10:59sudomani made a wiki account. could i please be given editing permission?
19:11:14AlexPWhat's your wiki name?
19:13:21AlexPOK, done
19:13:21 Quit flydutch (Quit: /* empty */)
19:13:31sudomanthanks ; )
19:14:52 Quit robin0800_ (Remote host closed the connection)
19:17:32 Join lyngaas [0] (
19:17:44 Quit Adub- (Quit: <Tobin> japanese men make hot trannies)
19:20:42 Quit shai (Quit: Leaving)
19:20:53 Join shai [0] (
19:20:55sudomani reported a bug a couple days ago where when installing rockbox for the first time on a 3G ipod, the screen boots up with the contrast all the way up so that it is black and almost impossible to see anything.
19:21:37sudomanas i mentioned in the report, this could be fixed by setting the default contrast to about 36
19:23:32AlexPJdGordon: What's the plan for the fms?
19:24:17 Quit Tomis (Ping timeout: 252 seconds)
19:27:46amiconnpixelma, sudoman: Yes, for ipod 1st..3rd Gen Menu+Play is reset, and Back+Next is diskmode
19:27:46 Quit HBK (Read error: Connection reset by peer)
19:28:12amiconnI didn't know the wiki stated the wrong combos. Iirc the manual is correct
19:28:22 Join HBK [0] (
19:28:51amiconnApple changed it with the clickwheels (4th Gen and later), since pressing the opposite sides is mechanically impossible on those
19:29:53sudomani updated the wiki for the subject
19:30:07sudomanwhat about ipod nanos?
19:30:25amiconnNano, mini etc are all clickwheel ipods as well
19:31:34 Quit shai (Quit: Leaving)
19:31:45 Join shai [0] (
19:31:57 Join stripwax_ [0] (
19:33:23 Quit stripwax (Ping timeout: 258 seconds)
19:40:08sudomanwould it be out of the question to have the option in rbutilqt to not check for new updates but install using whatever cached information it has? this could be useful if someone needs to reinstall the firmware but doesn't have an internet connection close by.
19:43:26 Join Tomis [0] (~Tomis@
19:49:16 Join toffe82 [0] (
19:50:31 Quit crashd (Read error: Operation timed out)
19:53:30domonokysudoman: rbutil already has a cache, but dont know if all code places really use it.
19:55:51domonokysudoman: there is also the "Offline mode" checkbox. Also not sure if it works for all.
19:56:55sudomandomonoky: thanks
19:59:31 Quit Jaykay (Ping timeout: 256 seconds)
20:08:50 Join anewuser [0] (anewuser@
20:08:53 Quit anewuser (Changing host)
20:08:53 Join anewuser [0] (anewuser@unaffiliated/anewuser)
20:16:16 Join Jaykay [0] (
20:17:10 Quit amiconn (Disconnected by services)
20:17:12 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
20:17:18 Quit pixelma (Disconnected by services)
20:17:20 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
20:17:36 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
20:17:37 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
20:17:38pamauryHEHEHE, look at that green delta ;)
20:18:40JdGordonAlexP: well, I need to figure out why hwcodec is broeknm
20:18:49JdGordonpixelma: did you get a chance to try changin that line?
20:18:55AlexPah, I didn't know it was
20:19:02AlexPpamaury: nice :)
20:19:11*AlexP misses the commit bot
20:19:59JdGordonis anyone else having trouble with svn hanging after a commit?
20:20:10AlexPIt isn't hanging, it is just *really* slow
20:20:21AlexPgive it a minute or two and it'll come back
20:20:22rasherYes, everyone. With CIA absent the commit hook times out before returning
20:20:59 Join lead [0] (
20:21:36AlexPJdGordon: The fms patch doesn't apply at the moment incidently (I guess multifont). I didn't bother to look at fixing it :)
20:22:01JdGordonI'm hoping to commit it this week... but depends on hwcodec testing
20:22:18JdGordonas much as I dont care about those targets, I don't exactly want to intentioally break them :p
20:23:00JdGordonpamaury: nice! shame its all ram though
20:24:41 Join KBH [0] (
20:26:44 Quit panni_ (Ping timeout: 252 seconds)
20:26:52 Quit HBK (Read error: Connection reset by peer)
20:30:53sudomanmy ipod dies at the battery level of 50%. is there any way to calibrate the indicator so that it's more accurate?
20:31:32 Join _zic [0] (
20:31:41TheSevenis there anyone around who's interested in participating in the nano4g work?
20:31:51TheSevenwe have a bootrom-level code execution exploit :-)
20:33:54 Join panni_ [0] (
20:35:09 Part jeffp
20:43:29 Quit stooo (Ping timeout: 240 seconds)
20:44:13 Join stooo [0] (
20:49:10amiconnlinuxstb: I wonder how to integrate the new mmx code in our build system
20:49:17linuxstbamiconn: gevaerts was asking the other day about using asm-optimisations in the sim (for his n900). So I guess using MMX optimisations are the same issue....
20:49:27amiconnDo we want to enable mmx permanently (for x86, that is)?
20:50:00amiconnOn x86_64, mmx is standard, so we can do that, but not all 32 bit x86's support it
20:50:14amiconnThat said, all non-stoneage x86's do
20:50:32rasherI'd say we can safely require mmx for the sim
20:51:14amiconnIntel supports it since Pentium MMX, AMD since the K6. Cyrix 6x86 does as well (it even implements a slightly extended variant)
20:51:38linuxstbAlthough maybe with the ability to disable it?
20:52:01amiconnThe way libdemac is built doesn't allow runtime detection (currently)
20:52:13rasherUnless it's fairly simple - I wouldn't spend much time on letting it be disabled
20:52:32amiconnIt's just a matter of passing -mmmx to gcc or not
20:52:33 Quit mc2739 (Ping timeout: 248 seconds)
20:52:43linuxstbgevaerts: How did you get on with enabling the ARM optimisations in the sim?
20:52:54linuxstbamiconn: Then I guess disabling it is just editing the Makefile to move that...
20:52:57amiconn(yes, 3 'm's)
20:53:02AlexPThen I'd enable it uncoditionally and just document that if people really want to run without they can
20:53:03 Join lude187 [0] (
20:53:16 Join gevaerts-phone [0] (~bc058851@rockbox/developer/gevaerts)
20:53:29 Quit lude187 (Client Quit)
20:54:03***Saving seen data "./dancer.seen"
20:54:25Buschelsudoman: you could try to set your battery capacity to a smaller value
20:54:31amiconnlinuxstb: Regarding the n900 - it uses an arm, right?
20:54:34 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
20:54:41linuxstbamiconn: Yes.
20:54:45gevaerts-phonelinuxstb: those ARM defines do the trick, together with enabling one or two more files in SOURCES
20:55:04gevaerts-phoneThat makes everything build, axcept ape
20:55:18amiconnWhat happens with ape?
20:55:46sudomanBuschel: right now it's at the smallest setting. (it says there's about 126h on a full charge, but that's a known problem) i think that setting doesn't affect the charge percent calculation anyways.
20:55:46linuxstbDid it give you a decent speed-up?
20:55:50gevaerts-phoneI'll tell you later, when I'm home
20:56:26gevaerts-phoneIt didn't fix the audio dropouts, I didn't do test_codec tests
20:57:31Buschelsudoman: what iPod do you use?
20:57:42sudomanBuschel: ipod 3G
20:58:53sudomanBuschel: it might be relevant that i installed a replacement battery a couple of years ago. i only get about 2.5 hours out of it at this point. (but that might not be from a completely full charge)
20:59:11AlexP% is from voltage IIRC
20:59:22AlexPSo if the battery is buggered that is probably it
20:59:36AlexPIt can't supply the current required so shuts down early
20:59:44sudomanBuschel: there's no way of telling with the rockbox firmware as it doesn't say when the battery is charging or not, but the apple firmware does
21:00:02sudomanAlexP: is there a way to fix that?
21:00:08Buschelsudoman: ah ok. The iPod 3G has no correct estimation implemented as there were no reliable battery runtime measurements availabe.
21:00:10AlexPreplace the battery
21:00:22 Quit tetsuo55 ()
21:00:23AlexPIf it is a dead battery
21:00:42sudomanAlexP: yeah, i might do that, or just get a 5.5G
21:00:54Buschelsudoman: We could easily implement the needed changes if you could do a battery bench with a good working (new?) battery.
21:00:58domonokythe % comes from the voltage AND the battery curve in the powermgmt-xxx.c files. so a different battery could also cause "low battery" shutdowns.
21:01:15AlexPah, OK
21:01:29AlexPIf the discharge curve is different
21:01:48domonokyor maybe the battery curve and min values were never calibrated for ipod 3g ?
21:01:54sudomanwould it not be useful to submit a curve with my current battery?
21:01:57AlexPcould be
21:02:12AlexPsudoman: Only if we know it is good
21:02:47 Nick KBH is now known as HBK (
21:02:50sudomanAlexP: would i need to get an apple-supplied battery, or could i use a souped-up one?
21:02:52 Join tetsuo__ [0] (
21:03:20*domonoky sees a: FIXME calibrate value for other 3g+ ipods in the relevant powermgmt-xxx.c file :-)
21:03:42AlexPhmmm :)
21:04:16AlexPsudoman: I'm not sure to be honest
21:04:24Buschelsudoman: it could also be a replacement battery. but we need to know the capacity in mAh
21:05:41domonokycurrent values for ipod3g: 3,37V = 0%. 3,38V is considered dangerious and 3,02V is shutdown.
21:05:45sudomanok. is this issue related?: when charging from a "dead" battery, the percent quickly goes up to 100%. does that mean the charging is terminated early, or that charging is continuing and that the calibration needs fixing?
21:06:23domonokythere is also a curve for charging, which is probably wrong too.
21:06:28Buschelsudoman: and of course you should perform the test following the standards (e.g. clean restart after full charge, looping a folder that is larger than the iPods RAM)
21:06:38sudomanhuh, my player shut down at 3.82V
21:07:14AlexPdomonoky: It'll also shutdown if it fails to spin up the disk right?
21:07:16sudoman^ (from my last test which didn't follow all of the standard rules)
21:07:24domonoky3.82V should show ~50% with the current battery curve.
21:07:42sudomandomonoky: that's correct
21:07:47 Quit gevaerts-phone (Quit: Page closed)
21:07:48 Quit moos (Quit: ChatZilla 0.9.86 [Firefox 3.6/20100115144158])
21:08:01domonokyAlexP: yes, that values are only pure software. We also shutdown if we cant spin up the HD, or if power goes just out.
21:08:04sudomanAlexP: it managed to write the value to disk one second after the previous entry, so it must not have been a disk issue
21:08:28AlexPI don't quite follow that sentence :)
21:09:37domonokysudoman: did it show the "low battery" warning before shutdown ?
21:09:56sudomanthe battery benchmark writes a new set of data points every minute. when the disk is accessed, it writes them to disk. my final value is listed as being one second later than the previous, which implies that it detected the low power and wrote to disk
21:10:04sudomandomonoky: no it did not
21:10:15AlexPsudoman: Does it?
21:10:22AlexPDoes battery nench work like that?
21:11:07pixelmaas far as I know battery_bench caches values in RAM until the next spinup (e.g. for rebuffering)
21:11:09sudomani was accessing a context menu or something. i guess it could have been a bug because i've experienced shutdowns while using the menus before. thought it was just a coincidence
21:11:18domonokysudoman: then it probably wasnt a shutdown based on the values from the battery curve.
21:12:14 Nick Guest57645 is now known as martian67 (
21:12:26sudoman*correction: it saves a new set of points to memory every minute, and writes to disk occasionally, according to the manual
21:12:29 Quit martian67 (Changing host)
21:12:29 Join martian67 [0] (~martian67@about/linux/regular/martian67)
21:12:35AlexPsudoman: indeed
21:13:04sudomandomonoky: was it a power-failure then?
21:13:05domonokysudoman: jup, if it would write directly to disk, the bench would not show normal behavior (much more power use)
21:13:27domonokysudoman: either that, or a bug :-)
21:13:48amiconngevaerts: Could you paste the output of gcc -E -dM - < /dev/null | cut -c 9- | sort on your n900 somewhere?
21:14:04sudomancould be a bug. i do remember having a bug/power-failure(?) at 40% once...
21:14:17Buschelsudoman: it could also have been a simple crash.
21:14:27domonokyif you expirienced shutdowns while using context menus before, it looks more like a bug.
21:14:45sudomanhow is a crash different than a bug?
21:15:40Buschelsudoman: could you please be so kind and perform a battery bench without touching your ipod? ;o) as soon as there is a full reasonable battery bench we can change the code to allow (more or less) proper estimation of runtime.
21:16:05sudomanBuschel: ok i'll do that ; )
21:18:34Buschelperfect :)
21:19:45amiconngevaerts: The output of uname -m would also be interesting
21:20:43 Quit Buschel ()
21:24:45sudomanfor the benchmark, should i charge with the rockbox firmware or the apple firmware?
21:25:14 Quit shai (Quit: Leaving)
21:25:29 Join shai [0] (
21:25:44linuxstbsudoman: Using the apple firmware is probably safer.
21:25:52linuxstb(and probably faster)
21:27:25pixelmafor the charging part
21:27:51sudomanpixelma: why is charging faster on the apple firmware?
21:27:58Unhelpfulamiconn: how hard would it be for the flashloader to handle a firmware with a compressed and uncompressed section? presumably the compressed portion would need a list of offsets and lengths, or it could be decompressed contiguously with start/end addresses for each section used to move things to their proper places.
21:28:54 Nick fxb is now known as fxb__ (
21:32:18amiconnQuite difficult... and it would mean that a full reflash is required
21:33:26amiconnI think it would better to make this part of crt0.S (which does the copying anyway, so it could uncompress instead of copying)
21:37:17amiconnThe question is whether it's worth it. We're talking about ~8KB of data (probably compressible to half of that)
21:38:35sudomanis it ok to switch on "hold" during the battery test?
21:40:41danopia_so where do you guys host SVN? sourceforge?
21:41:12danopia_my commit announcer works with sourceforge now, it acts as a mailing list subscriber
21:41:35 Join freddyb [0] (
21:44:44 Quit tetsuo__ ()
21:45:09danopia_that makes it a little harder :P
21:46:55 Join checker [0] (
21:48:12domonokydanopia_: but we ofcourse also have a commit mailing list, so you could just subscribe to that.
21:48:43danopia_domonoky, where is that, and if it works, are you interested in me putting an announcing bot in here (at least until CIA lives)?
21:49:34 Quit shai (Quit: Leaving)
21:49:42 Join shai [0] (
21:50:38 Quit bertrik (Read error: Connection reset by peer)
21:50:44domonokydanopia_: a interim commit message bot till CIA is back would be cool.
21:50:54 Quit shai (Read error: Connection reset by peer)
21:51:00 Join shai [0] (
21:52:43tomersdoes anybody knows how to debug a plugin using DEBUGF? Should I reconf it with logf enabled?
21:55:03*domonoky thinks DEBUGF only works on sims (and maybe needs a debugbuild)., for target debug, use logf with a logf enabled build.
21:58:58rasherTechnically it also works for targets with a gdb stub, doesn't it?
21:59:56amiconnIt does
22:00:17 Quit Jaykay (Ping timeout: 248 seconds)
22:06:57JdGordonpixelma: ping?
22:07:39 Quit utchybann (Quit: WeeChat 0.3.0)
22:09:30 Quit stoffel (Read error: Connection reset by peer)
22:13:23 Quit S_a_i_n_t_ (Read error: Connection reset by peer)
22:14:39 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
22:15:11 Quit GeekShado_ (Read error: Connection reset by peer)
22:16:23 Join bertrik [0] (
22:16:58 Join S_a_i_n_t [0] (S_a_i_n_t@
22:17:53 Quit S_a_i_n_t (Client Quit)
22:20:32Unhelpfulamiconn: and we're over by how much, now? :/
22:21:06 Join Utchybann [0] (
22:24:23 Join Jaykay [0] (
22:24:35 Quit n17ikh (Ping timeout: 256 seconds)
22:30:07gevaertsamiconn: the build environment is in scratchbox, i.e. not actually on the phone (I gave up on that, too impractical)
22:30:14 Join n17ikh [0] (
22:30:22gevaertsamiconn: gcc -E -dM - < /dev/null | cut -c 9- | sort output :
22:30:43gevaertsuname -m says "arm" in scratchbox, and armv7l on the actual phone
22:31:11 Quit Jaykay (Quit: ChatZilla 0.9.86 [Firefox 3.6/20100115144158])
22:33:03 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
22:33:09gevaertsape compilation fails with
22:33:33gevaertsThis is with -DCPU_ARM -DARM_ARCH=6 added to EXTRA_DEFINES
22:39:41JdGordonanyone think there is any point trying to optimise the skin displayer code?
22:42:33gevaertsin what way?
22:43:07JdGordonnot really sure, but its called pretty often and if its worth getting 1% out of the codecs it might just be worth getting the same from there
22:43:26JdGordonI'm sure there is plenty of room for improvement
22:45:08amiconngevaerts: Hmm, strange. Would it be possible to get to see this intermediate .s file somehow?
22:45:32 Quit lyngaas (Ping timeout: 256 seconds)
22:47:40amiconnHmm, eabi?
22:48:05 Quit tomers (Ping timeout: 252 seconds)
22:48:06gevaertsI think so, yes
22:49:02gevaertsamiconn: adding -S to the gcc line gives me gevaerts/output.s">
22:54:05***Saving seen data "./dancer.seen"
22:55:50amiconnHrmm. It seems "5.35.1 Size of an asm" is the problem
22:57:14amiconnI have no idea how to solve this, apart from removing all usage of .rept and manually replicating all this code in vector_math??_*.h
22:58:34gevaertsit's not actually a real problem right now I think
22:59:09 Quit bertrik (Quit: De groeten)
22:59:31gevaertsI mean, it breaks compilation of ape in this particular weird setup, but there are more important bugs to look at first before I can actually start thinking about actually playing ape files
23:01:27LloreanIs part of the website broken? I can't get to flyspray.
23:01:49gevaertsI can
23:02:00 Join saratoga [0] (
23:02:12saratogai wonder if the dircache improvements help playback on the c200v2
23:02:19saratogathe added memory might make a difference
23:02:33saratoga30KB is a large fraction of the entire audio buffer on that player
23:02:40LloreanCan someone tell FS #11011 that while it may not be the reset combo for his iPod, it certainly is for the Nano?
23:02:46*Llorean isn't sure which is the reset combo for the Mini.
23:03:29JdGordonmenu+select is a hard reset
23:04:42 Quit _zic (Remote host closed the connection)
23:05:32sudomanLlorean: i wrote that. i guess some nanos have click-wheel and some don't?
23:06:03 Join m3dlg [0] (
23:06:08LloreanAll nanos use Menu+Select to reboot
23:06:25gevaertsas far as I remember, mini 2g is the same as video
23:06:33sudomanoops. sorry. i got the nanos and minis backwards
23:06:40LloreanI though the only ones that don't use Menu+Select are ones that have buttons independent of the wheel.
23:06:50pixelmasudoman: it's been answered to you
23:06:52sudomanit only applies to 1G - 3G ipods
23:07:23sudomanpixelma: sorry, i don't understand
23:07:27linuxstbsudoman: All minis have a clickwheel.
23:07:31mtI thought pamaury meant 30Kbits because of (Kb) ?
23:08:16linuxstbsudoman: Are you perhaps confusing "1st/2nd gen mini" wiith "1st/2nd/3rd gen ipods" ?
23:08:46sudomanlinuxstb: not sure, i think i just got my memory backwards for the minis and nanos
23:08:56linuxstbsudoman: But the minis and nanos are the same
23:09:01pixelmasudoman: and the following minutes
23:09:04linuxstb(in terms of key combinations)
23:09:12sudomani mean that i got them both wrong
23:10:04Unhelpfulamiconn: i was mostly thinking "loader has a decompressor already, maybe using that would save some more space"... perhaps a simple header could indicate the offset and length to the compressed portion, and the loader could always decompress it to a fixed, and contiguous, place in DRAM (which would be given as the load address for the various things that we might compress)?
23:13:23pixelmaUnhelpful: how much it is "over" depends on the target, I guess for the FM-Recorder it is quite much already. IIRC rombox was broken on those about 3 years ago (or so) with the addition of cuesheet support and that endianness independent handling of tagcache files
23:14:33 Quit martian67 (Remote host closed the connection)
23:15:50 Quit anewuser (Quit: =ooo ϢINTER ϾHIP 5iVE is OOON!!)
23:16:16pixelmawow, cuesheet support was added *exactly* 3 years ago... and I think I remember correctly that this broke rombox on the FM-Recorders
23:16:49JdGordonwas it me or nicoP who got the commit in for cuesheets?
23:17:10mtsaratoga: atrac3 is still not realtime on cf targets with little iram
23:17:37UnhelpfulJdGordon: are you making notches in your "modern" player for every "ancient" player on which you break rombox? ;)
23:17:50pixelmaJdGordon: Nico_P, but the commit message reads "patch by JdGordon"
23:18:19JdGordonsure, go ahead and blame me!
23:18:24*gevaerts blames JdGordon
23:18:38JdGordonUnhelpful: not on my dap... the side of my cell's wall though :)
23:18:55 Join martian67 [0] (~martian67@about/linux/regular/martian67)
23:19:34 Quit Llorean (Quit: Leaving.)
23:19:36 Quit martian67 (Remote host closed the connection)
23:19:50pixelmaendianness independency of tagcache was committed February 13th 2007 and added a lot too
23:21:31Unhelpfulpixelma: easy fix, make the on-disk endianness of the tagcache same as that of sh1 ;)
23:22:09pixelmaJdGordon: "patch by JdGordon and me" ... you're not the only one to blame then
23:22:31Unhelpfulonly good for a half-notch, then?
23:23:02JdGordonpixelma: have you got time to play with the fm patch?
23:24:21JdGordonI would really like to get this in soon
23:24:27 Join lyngaas [0] (
23:25:14amiconnUnhelpful: The default endianness for database is always equal to target endianness. The code can handle both, this is to support external database creation
23:25:33pixelmaJdGordon: no. One thing I tried though is a peakmeter in the sbs and it works in every screen but lists. Looks like it doesn't update in lists - thought it was worth checking if it works in the radio and recording screen in general
23:26:06amiconnWe *could* disable that for lowmem, but then these targets profit most from external creation, since on target it's slow (not unbearably slow though)
23:29:04pixelmaJdGordon: do you want to get it in so quickly to break the OndioSP's rombox for good? ;\
23:29:20JdGordonmy evil plot has been found out!
23:30:00 Join Buschel [0] (
23:30:05JdGordonpixelma: can you try commenting out apps/recorder/radio.c lines 991-995 and replace them with splash(0,"something"); and see if you ever see that splash?
23:30:46JdGordonand re rombox, multifont added hardly anything to bin size, it just shows how close it has been to being killed of for good for ages
23:31:23pixelmaok, you need some patience though, testing on the Ondio is a bit tedious :/
23:31:53 Quit checker (Quit: CGI:IRC (EOF))
23:32:25JdGordonwanna setup ssh and a robot and a webcam so I can play from here? :)
23:32:58pixelmayou won't test much anyways :P
23:33:21 Join martian67 [0] (~martian67@about/linux/regular/martian67)
23:34:38 Quit martian67 (Remote host closed the connection)
23:34:41pixelmadid you try checkwps with the radio screen? Unfortunately checkwps doesn't build for me and since it fails in a file that is untouched by the patch, I think it is cygwin related
23:35:10 Join martian67 [0] (~martian67@about/linux/regular/martian67)
23:35:20Buschelmt: is mdctexp faster for atrac on coldfire as well?
23:36:14JdGordonpixelma: no, i'll try that now
23:36:40mtBuschel: I'm not sure, the measurements were all done on arm targets.
23:37:09domonokyJdGordon: and when you commit the fms patch, remember that the themesite needs updates for that. (since my newest addition it wont a accept a fms without changes, because thats a unknown theme cfg value )
23:37:17Buschelmt: somebody should support you with some measurements...
23:37:32JdGordondomonoky: arg, is that a recent change?
23:37:39JdGordonI thought it didnt go through the .cfg?
23:38:23domonokyJdGordon: yes, today. But you can add new allowed settings via the admin interface. (still needs changes to actually check a fms with checkwps)
23:38:24JdGordonok, the hardcoded fms passes checkwps
23:38:35 Join akur [0] (
23:38:50JdGordonI'm not an admin on the theme site
23:39:06pixelmathe one with %pm in the conditional?
23:39:11Buschelmt: atrac's iqmf synthesis filter is nearly fully optimized on arm now. without dropping precision and/or dropping coefficients there is not much left to do.
23:39:31domonokyJdGordon: ask rasher to give you admin access :-)
23:40:14rasherdomonoky: surely you have access to that as well - or are you just bothering me because I'm responsible for the current lack of admin interface for this? :)
23:40:17JdGordonits more fun to annoy others to fix ti
23:40:23mtBuschel: Wouldn't say no to cf measurements of course :)
23:41:20domonokyrasher: no i cant do that. my login doesnt have write access to the db, i think
23:41:32mtBuschel: Do you think there's something else that could be optimized ?
23:41:55JdGordonarg, I reckon there is enough duplicated code for the skin init/display stuff that we could get the ondio rombox going again, but that will only give us another few weeks... so what's the point?
23:41:59rasherdomonoky: ah yes, the sqlite db had the wrong permissions at one point - it should be fixed now
23:42:12domonokyah, then i can do it too :-)
23:42:13rasherdomonoky: this also meant that checkwps wasn't able to update the db from cron
23:42:37pixelmaJdGordon: the point is that it's always nice to save bin size/ RAM usage
23:42:53JdGordonpixelma: the previous tests, you were just getting a blank fm area? the statusbar kept updating?
23:43:01Buschelmt: i had some questions in the area of joint stereo coding. especially the INTERPOLATE macros could be optimized and the div16 calls can be avoided via lookup tables...
23:43:03mtBuschel: I tried the the reversal of spectral coefficients of odd bands outside the imlt, but the gain was embarrassing; 0.16% on arm :-)
23:43:47pixelmaJdGordon: statusbar was displayed, yes. I didn't pay much attention if it was updated or not though
23:44:23JdGordonyeah, ok, if you could try the splash change and let me know if it ever shows up that would be great
23:45:25Buschelmt: afterwards i saw that i have no file with joint stereo at all :/
23:45:53Buschelmt: nothing to optimize or verify...
23:46:04mtBuschel: Yes, was just going to point that out
23:46:45Buschelmt: nevertheless: is the calcualtion in fastSqrt(...) correct?
23:47:03Buschelmt: or better: the way it is used?
23:47:56 Join asjsafjoa [0] (~47eca4cc@gateway/web/freenode/x-etbwqagsehrkkwrh)
23:48:46Buschelmt: e.g. when ind==7 -> ch[0]=ONE_16 ch[1]=ONE_16. but when using ch[0]=ONE_16 withni the fastSqrt(...) formula the result is ch[1]=sqrt(ONE-16).
23:48:48mtBuschel: I did some tests against the floating point decoder back when I was still converting the codec to fixed point, and it was correct.
23:49:10Buschelmt: could you test with joint stereo files?
23:49:29pixelmaJdGordon: and if it really is duplicated code, a cleanup will help maintaining the code a lot
23:49:44mtBuschel: Guess I have to now ! .. I'll look for some.
23:49:45pixelmaif you remove it, I mean
23:50:24Buschelmt: we should make a lookup table for ch[0] and ch[1] to avoid the calculations.
23:50:27JdGordonno, well the case can be made that it makes the code harder to read
23:50:35Buschelmt: if you find some, give me a message ;o)
23:50:52mtSure. :)
23:52:50pamaurymt: I don't unerstand why the build_dev tables show 15Kb. On my computer I had a 30Kb augmentation irrc. But perhaps I made a mistake
23:54:18pixelmathe delta table takes bin size and RAM usage into account (and takes the average)
23:54:22 Quit asjsafjoa (Quit: Page closed)
23:54:48mtpamaury: I don't know about that - I was just asking because saratoga said 30KB, while your message said 30Kb i.e < 4 KB. I was just confused whether you actually meant bits or bytes. :)
23:55:08 Join fml [0] (
23:55:43pamaurymt: I meant bytes
23:55:48pixelmapamaury: if you hove over a value in the table you can see both in the tooltip
23:56:08fmlJdGordon: isn't there a signedness mismatch in buffer in skin_font and buffer_start in font
23:56:33pamaurymt: why the table show: bin: -420, ram: -30xxx and a global -15000 ?
23:56:42B4gderpamaury: average
23:56:50B4gderas pixelma explained
23:57:23*gevaerts thinks that average is a bit meaningless there
23:57:27pamauryAh, I didn't see pixelma comment :) I've been afk for the last hours
23:57:35rasherMaybe it should just be the sum?
23:57:39*pamaury agrees with gevaerts
23:57:50B4gdersum might be better, yes
23:57:51pixelmapamaury: it was just below your statement
23:57:52rasherOr else it should be a score, rather than a delta
23:58:11gevaertssum is also wrong. ram includes bin
23:58:48pamaurytake max ?

Previous day | Next day