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 2012-08-04

00:01:19 Quit benedikt93 (Quit: Bye ;))
00:01:22 Quit Totalled (Ping timeout: 245 seconds)
00:05:36 Quit n1s (Ping timeout: 240 seconds)
00:07:15 Join perrikwp_ [0] (~quassel@cpe-024-163-024-033.triad.res.rr.com)
00:10:16 Quit perrikwp (Ping timeout: 264 seconds)
00:12:17 Join amayer [0] (~amayer@h190.51.21.98.dynamic.ip.windstream.net)
00:16:42 Join Totalled [0] (~Totalled@c-98-245-9-211.hsd1.co.comcast.net)
00:21:28 Quit TheDarkPirate (Quit: Leaving.)
00:23:28 Quit Totalled (Ping timeout: 264 seconds)
00:23:52 Quit kevku (Quit: KVIrc 4.2.0 Equilibrium http://www.kvirc.net/)
00:26:57 Join perrikwp [0] (~quassel@cpe-024-163-024-033.triad.res.rr.com)
00:27:19 Join Totalled [0] (~Totalled@c-98-245-9-211.hsd1.co.comcast.net)
00:29:55 Quit perrikwp_ (Ping timeout: 272 seconds)
00:54:29 Quit eckoit (Quit: eckoit)
00:55:10 Join perrikwp_ [0] (~quassel@cpe-024-163-024-033.triad.res.rr.com)
00:57:41 Quit perrikwp (Ping timeout: 246 seconds)
01:00
01:01:11 Quit mgottschlag (Ping timeout: 246 seconds)
01:01:41bertrikfreqmod, something goes wrong if I try to seek to granule 0 with opus
01:07:58 Quit ender (Quit: When one person suffers from a delusion it is called insanity. When many people suffer from a delusion it is called religion. -- Robert Pirsig)
01:15:21 Join perrikwp [0] (~quassel@cpe-024-163-024-033.triad.res.rr.com)
01:17:48 Quit perrikwp_ (Ping timeout: 246 seconds)
01:18:04 Quit Prodicus (Ping timeout: 248 seconds)
01:18:54 Join Prodicus [0] (~chatzilla@69.169.144.239.provo.static.broadweavenetworks.net)
01:24:10amayerdo i need any of the add-ons to use git with rockbox code?
01:24:44bertrikadd-ons?
01:25:30bertrikyou need the git-core package, and I think that has all the git stuff you need
01:25:33amayeri looked it up in the ubuntu software center and it has:
01:25:34amayergit-arch
01:25:36amayergit-cvs
01:25:38amayergit-daemon-run
01:26:45amayerso just git not any of the other stuff?
01:26:53bertrikright
01:27:32amayerif you cant tell im new to the git thing.
01:27:34amayerat my work we just use a central server. and yell across the hall to see if anyone is working on the file we need
01:27:51amayerbut then again it is only 4 of use working on the stuff
01:28:30bertrika revision control system is useful even if you're the only person working on something
01:29:00amayerim not saying its not useful. i was just saying i have no experience with the like at all
01:29:04 Join perrikwp_ [0] (~quassel@cpe-024-163-024-033.triad.res.rr.com)
01:29:11amayerim reading the wiki page tho
01:29:18bertrikeven subversion is better than nothing at all
01:29:36 Quit perrikwp (Ping timeout: 240 seconds)
01:34:23amayerhaha we have a nas that does backups
01:34:41 Quit mikroflops (Ping timeout: 245 seconds)
01:35:27amayerwe are a pretty small firm. most of the software and web development falls on my lap... and the only thing my boss knows about computers is that his is running windows and mine runs linux(but he doesnt know what that means either)
01:35:34*gevaerts would like to point out that discussion of why VCSes and backup tools are different things is a #rockbox-community matter
01:35:54amayergevaerts: sorry
01:36:39 Join mikroflops [0] (~yogurt@h-34-21.a238.priv.bahnhof.se)
01:42:23***Saving seen data "./dancer.seen"
01:47:58 Quit perrikwp_ (Read error: Operation timed out)
01:48:07 Join perrikwp [0] (~quassel@cpe-024-163-024-033.triad.res.rr.com)
01:50:54 Join saratoga [0] (98032941@gateway/web/freenode/ip.152.3.41.65)
01:51:22saratoganew resampler is pretty much ready to go aside from optimization
01:51:31saratogaso 48k with opus shouldn't be an issue
01:56:53 Quit pamaury (Remote host closed the connection)
01:57:15gmaxwellsaratoga: does the hardware rockbox runs on not output 48KHz?
01:57:36saratogagmaxwell: most devices do, but we operate our DSP pipeline at 44.1k
01:57:45gmaxwellBummer.
01:58:08saratogavery little music actually uses 48k, so we never bothered to have good quality for it
01:58:36gmaxwellHopefully none of thoe devices have AC97 codecs. :)
02:00
02:00:23saratogapretty much the same thing
02:01:15ProdicusHow much work would be involved in getting native 48k playback? What stuff would have to be rewritten?
02:01:38saratogaProdicus: we decided it wasn't worthwhile
02:02:07saratogaor at least not supporting multiple sampling rates
02:02:13gevaertsI'd expect that fully switching to 48k shoudln't be too tricky, but that doesn't really solve anything
02:02:30saratogai guess you could do a compile that was always 48k without a huge amount of work
02:02:43saratogabut yeah everything else gets resampled
02:04:05ProdicusInteresting. Is your current resampler particularly efficient for small fractions i.e. would 24k, 16k playback also benefit from a version compiled for 48k output?
02:04:33saratogaright now its just linear interpolation, so the ratio doesn't really matter
02:05:05Prodicusoof.
02:05:21gmaxwellugh
02:05:29saratoga g#304 is hermite, which also probably doesn't care too much (although i guess all interpolations work better going from very high sampling rates to much lower ones)
02:05:31fs-bluebotGerrit review #304 at http://gerrit.rockbox.org/r/304 : Introduce new hermite polynomial resampler. by Michael Giacomelli (changes/04/304/1)
02:05:42gmaxwellwith 48k to 44.1k... directly?
02:06:07gmaxwell17:00 < saratoga> pretty much the same thing
02:06:14gmaxwellwhat is pretty much the same thing?
02:06:49saratogathe DACs we use are basically the same as AC97, although obviously with I2S or whatever as the interface instead of PCI
02:07:14gevaertsThe resampler is an old issue. I'd say too many of us have tin ears, so we haven't been motivated to fix it
02:07:15saratogaor does AC97 not use PCI?
02:07:48gmaxwellsaratoga: well virtually (?) all AC97 dacs are horrible at 44.1k because they internally run at some multiple of 48k and there is a very low precision linear resampler for other rates.
02:08:12saratogagmaxwell: thats just a software thing, you can clock a DAC at whatever you want if you don't care about windows compatibility
02:08:26 Join Syconaut^ [0] (viper@c-4dfd72d5.162-1-64736c10.cust.bredbandsbolaget.se)
02:09:24 Join perrikwp_ [0] (~quassel@cpe-024-163-024-033.triad.res.rr.com)
02:09:50gmaxwellsaratoga: if you have an oscillator which can be divided down the a relevant rate... but it's not like 48k and 44100 are both common denominators for many possible clocks.
02:10:27saratogagmaxwell: the oscillator depends on the system, not the DAC, and virtually all systems have dividers for both sample rates
02:10:55saratogasince we can control the PLL and dividers we usually don't have much trouble, although sometimes it can be tricky to get the clock just right without breaking something else
02:11:08gmaxwellOkay thats good then!
02:11:41saratogabut yeah, an AC97 DAC and a mp3 player DAC are not much different aside from the digital interface
02:12:01saratogaactually there is very little difference between any modern DACs
02:12:03 Quit Syconaut (Ping timeout: 248 seconds)
02:12:24 Quit perrikwp (Ping timeout: 246 seconds)
02:13:34gmaxwellSure, I was just under the impression that the crap linear resampling and the lack of ability to run 44.1k was something specific to the interface used on AC97 since its the only place where I've seen it be an obvious issue on PC hardware (/linux drivers).
02:14:20saratogaAC97 defines some chipset level interface that requires 48k (or maybe some other sampling rates)
02:15:07saratogabut if you're not hooked up to an intel chipset, thats not a problem
02:15:45 Join Prodicus_ [0] (~chatzilla@69.169.144.239.provo.static.broadweavenetworks.net)
02:15:54 Quit Prodicus (Ping timeout: 246 seconds)
02:16:04 Nick Prodicus_ is now known as Prodicus (~chatzilla@69.169.144.239.provo.static.broadweavenetworks.net)
02:17:57saratogai guess the downside of not using AC97 is that we have to write drivers for each and every DAC
02:21:33saratogai'm really not understanding the ASM code we have for the linear resampler
02:21:52saratogaits somehow gigantic even though its just taking the weighted average of two numbers?
02:24:58ProdicusI've been thinking that once the rockbox opus support is solid and the opus experimental vbr etc encoder stuff is released I'll reencode my entire collection in opus and flash my clip zip. So I would personally be interested in a 48k-native version as I'd no longer be worried about 22/44 kHz playback.
02:27:37 Quit bertrik (Ping timeout: 246 seconds)
02:28:40amayerdo different media types use different kHz or is it a hardware thing?
02:28:57saratogaopus uses 48k internally
02:29:06saratogaall other formats support various sampling rates
02:29:18 Quit lebellium (Quit: ChatZilla 0.9.88.2 [Firefox 15.0/20120724191344])
02:32:19amayerso opus is only 48k?
02:32:52saratogayeah
02:34:29ProdicusOpus encoding and playback tools can deal with any sample rates. But internally it uses 48k (and a handful of integer divisors) and requires resampling for other output rates.
02:36:07ProdicusSo by avoiding the final-stage resampler, 48k playback would be a bit more efficient and avoid some distortion.
02:36:49preglowis this about us still using 44k for everything internally? :>
02:37:00saratogayeah
02:37:03preglowblame coldfire
02:37:08preglowno
02:37:10preglowblame iriver!
02:37:34saratogaits not just that, its too difficult to switch sample rates and would break too much
02:37:46preglowtoo difficult in what way?
02:37:57saratogareclocking during playback without glitching is not simple
02:38:02preglowduring playback?
02:38:06preglowwhy would you do that?
02:38:31saratogaif you don't want to resample then sooner or later you'll have to reclock
02:38:57preglowwell sure, but you won't be doing it during playback
02:39:17Prodicusbetween tracks/albums instead
02:39:21saratogastopping playback => really long glitch
02:39:26preglownot really long
02:39:27preglowlongish
02:39:42preglowlike what? 250ms max?
02:39:51gmaxwell17:36 < Prodicus> So by avoiding the final-stage resampler, 48k playback would be a bit more efficient and avoid some distortion.
02:39:58saratogaits less annoying to just properly resample
02:40:12saratogano glitch, and good resampling isn't actually that expensive
02:40:15gmaxwellnormally you don't have any reason to worry about distortion from a competently written resampler.. but linear isn't competent. :( Fast though.
02:40:35preglowlinear is a throwback from when doing better meant having choppy playback
02:40:44Prodicusgmaxwell: exactly.
02:40:49gmaxwelleven fairly good resamplers can be pretty fast.
02:41:05preglowsure
02:41:06preglowwrite it
02:41:22saratogai did :)
02:41:32preglowpatch up?
02:41:49saratogahttp://gerrit.rockbox.org/r/#/c/304/
02:42:04preglowawesome
02:42:29saratogawas thinking about trying a higher order polynomial as well
02:42:34saratogabut this is still enormously better
02:42:52preglowthat looks awfully much like pure hermite
02:42:59saratogait is
02:43:04preglowwhat happened to the upsampling?
02:43:14saratogaoh haven't done that yet
02:43:19preglowbut it's in the works?
02:43:28preglowhermite spline is an improvement, but not by much
02:43:31saratogawell i made a filter in matlab, but thats it :)
02:44:06preglowwell, i can't talk, all i did was start reading the code of the speex resampler
02:44:18saratogawhats it do?
02:44:29preglowjean-marc spoke so convincingly of it to me i felt i had to have a look at it
02:44:45preglowit's fir based
02:45:00preglowas far as i know i think it's a variant of the jos resampler
02:45:03saratogafixed ratios only then?
02:45:06preglowbut it kinda does it two ways
02:45:11preglowno, i don't think so
02:45:21gmaxwellIt has code to update the filters.
02:45:24saratogais it in our code?
02:45:27preglownope
02:45:29preglowi think i left it out
02:45:41preglowit not being needed to compile speex
02:45:58gmaxwellDon't ask me how that works. It's pretty good though... might not be fast enough in fixed point for you though. It's also not terribly large.
02:46:11preglowgmaxwell: the cheap mode would work good for our faster targets
02:47:06saratogahttp://git.rockbox.org/?p=rockbox.git;a=blob;f=lib/rbcodec/codecs/libspeex/resample.c;h=65dfef28a3dc684e56bb42e3157714490d5d13c5;hb=HEAD
02:47:13preglowthat would be it :>
02:47:27preglowit's basically untouched since that rev
02:47:30preglowso that's it
02:47:34Prodicusre gmaxwell's comment 6min ago: Sometimes a moderately competent resampler can still be noticeable for other reasons. For instance I've found that using SoX's resampler rather than allowing LAME to use its internal polyphase resampler gives a noticeable quality difference when you're downsampling to 16k. Probably because of the wider passband and flatter response in the passband.
02:48:19gmaxwellProdicus: well it matters a lot more for lower frequencies than it does near 44.1k/48k.
02:48:23preglowProdicus: in who's favour?
02:48:29gmaxwelllittle differences in the passband can be noticable.
02:48:30Prodicusgmaxwell:sure
02:48:46preglowi doubt they're very different in most of the passband
02:49:02gmaxwellpreglow: right but the cutoff is audiable.
02:49:07preglowyep
02:49:13ProdicusSoX's. Its default resampler is rather high quality and has a passband of 95%
02:49:15preglowthat's usually where you hear the difference
02:49:21preglowProdicus: yep, as expected
02:49:27saratogaheh speex also uses kissfft
02:49:33preglowsox's resampler is of the usual overkill sinc type
02:49:39saratogadid you ever optimize speex's ffts?
02:49:46preglowspeex doesn't use any for decode
02:49:50preglownor encode, i think
02:50:03gmaxwellOh it uses fft in encode for sure, at least now.
02:50:08gmaxwellBut not for decode.
02:50:25preglowdecode is pretty much pure filterbank fidgetry
02:50:30gmaxwellCurrent speex has a toy version of the vorbis psymodel for noise shaping.
02:50:52saratogaah ok
02:51:12preglowgmaxwell: there's no reason for it to use ffts for encode, really, celp encoders usually never did, at least
02:51:27preglowsounds more like an echo canceller thing
02:53:03preglowthe psy model certainly would need it
02:53:50gmaxwellThe noice canceler and AEC use it too.
02:53:59preglowi love giving opinions without doing a grep :>
02:54:22gmaxwellwell, I haven't looked at the speex code in years either. so ::shrugs::
02:54:49preglowhaven't looked at it since i put in the decoder, but i never looked much at the encoder then either
02:55:18preglowjust talking off a vague understanding of celp encoders in general
02:56:42gmaxwellYou'll have fun looking at opus' linear predictive coder hah.
02:56:46preglowbut jm knows what he's about, having a look at the resampler would be clever
02:57:01preglowwhich part of opus? i've had a look at celt
02:57:13 Part amayer
02:57:13preglowthe speech coder is basicaly silk, afaik
02:57:41saratogais his resampler fast?
02:57:46gmaxwellYes, silk is a acelp codec which spent far too long being overdesigned. :)
02:57:50saratogaIMO we just need one good enough for portable use
02:57:56preglowsaratoga: he claimed it was meant for embedded platforms
02:57:58 Join amayer [0] (~amayer@h190.51.21.98.dynamic.ip.windstream.net)
02:57:58gmaxwellsaratoga: it can be, variable quality.
02:58:13preglowsaratoga: also that he made it just so resamplers like mine would never have to be made again :>
02:58:26preglowi didn't buy it completely at the time, but those were different times
02:58:26saratogaok i'll have to look at it
02:58:30saratogabut so much code
02:58:38Prodicusdoesn't opus include an updated copy of the speex resampler code?
02:58:45gmaxwellsaratoga: the silk in opus is substantially change from the original silk— the overall design is the same but all the details changed (though there still are some dumb things in it, oh well)
02:58:50gmaxwellProdicus: opus-tools.
02:58:55preglowProdicus: it shouldn't, speex is no part of opus
02:59:23preglowProdicus: ah, resampler, nvm
02:59:28Prodicusnot because it uses speex, because xiph had working code and reused it.
02:59:34gmaxwellProdicus: but the only 'update' is fixing a multichannel bug with weird sampling rates that have enormous prime factors.
02:59:40Prodicusah.
03:00
03:00:02preglowsaratoga: the core is quite simple, really
03:00:20preglowsaratoga: possibly not simple to optimize...
03:00:26gmaxwellmulti also means 2 so it might be relevant for you, if you care about 22047 hz 0_o.
03:01:14gmaxwellthat it's a fir resampler makes the critical path fairly easy to optimize.
03:01:28preglowi'd start with porting the _direct one anyway
03:02:45gmaxwellwell direct uses a fairly large amount of memory for 44.1k/48k. I'm using direct for that in opus-tools now, but it's not the default for the version in the speex code for that ratio.
03:02:59preglowyeah, it doesn't perform too well for that
03:02:59gmaxwellcertantly easy to make fast, it's just a multiply add loop.
03:03:22preglowbut few things do unless you do it a bit more elaborately
03:03:25gmaxwellpreglow: well on x86 desktops direct is a bunch faster even for 44.1k/48k which is why opus-tools uses it now.
03:03:55preglowwhat, resampling is a concern on x86 desktops? :P
03:04:04gmaxwell(just because it has fewer multiplies.. of course this is for quality 5.. q=1 probably not :) )
03:04:43gmaxwellpreglow: upsampling 44.1k -> 48k for input.
03:05:13gmaxwellOpus-tools also preserves the original rate by default on decode, and exact sample length— just because that the expectation for a file encode/decode tool.
03:06:02gmaxwellI didn't think the performance really mattered but I got people asking me to add crazy features because they wanted to use the sox resampler because it was faster.
03:07:03gmaxwellSo there was twiddling to get it faster to make the crazy feature requests go away. :)
03:08:35saratogaso basically its 4x oversampled lagrange using a 32 tap oversampling filter
03:08:52saratogafor 82.3% cutoff ( ~60 dB stop)
03:08:54 Quit Rower85 (Quit: Hmmm...)
03:27:45 Join eckoit [0] (~ryan@50.65.10.24)
03:42:27***Saving seen data "./dancer.seen"
04:00
04:02:56 Join zz_TheSphinX^ [0] (~briehl@p5B321732.dip.t-dialin.net)
04:06:22 Quit TheSphin_ (Ping timeout: 245 seconds)
04:09:44 Quit TheSeven (Disconnected by services)
04:09:53 Join [7] [0] (~quassel@rockbox/developer/TheSeven)
04:17:56amayerwhat is your email used for in git?
04:19:22 Quit Totalled (Ping timeout: 244 seconds)
04:27:36 Join pixelma_ [0] (pixelma@rockbox/staff/pixelma)
04:27:37 Quit pixelma (Disconnected by services)
04:27:41 Quit amiconn (Disconnected by services)
04:27:41 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:27:46 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:29:13saratogait shows up on all your commits
04:29:17saratogaits basically your username
04:29:19saratogaamayer:
04:29:34 Quit saratoga (Quit: Page closed)
04:29:58amayersaratoga: so we wont get emails everytime there is a commit or something?
04:30:25amayerits just like a signature?
04:31:31 Quit Thra11 (Ping timeout: 246 seconds)
04:46:18 Quit eckoit (Quit: eckoit)
05:00
05:11:01amayerhow do i compile the uisimulator?
05:11:26amayeri just set up git(no ssh yet) and would like to test a theme i am making
05:11:28amayer(baby steps)
05:11:57amayerim on linux btw
05:31:38 Join Totalled [0] (~Totalled@c-98-245-9-211.hsd1.co.comcast.net)
05:42:29***Saving seen data "./dancer.seen"
05:55:12 Join Amqui [0] (~a@d50-99-201-71.abhsia.telus.net)
05:55:18Amquigood evening
06:00
06:04:50amayerAmqui: hello
06:17:24 Quit scorche (Ping timeout: 248 seconds)
06:23:59 Join scorche [0] (~scorche@rockbox/administrator/scorche)
06:26:35 Nick perrikwp_ is now known as perrikwp (~quassel@cpe-024-163-024-033.triad.res.rr.com)
07:00
07:03:36 Join kevku [0] (x@heaaqi4aafadxgaamaafaadyaby.dyn.reverse.name)
07:05:17 Quit Prodicus (Quit: ChatZilla 0.9.88.2 [Firefox 10.0.2/20120216092139])
07:10:14 Part amayer
07:42:32***Saving seen data "./dancer.seen"
08:00
08:39:01 Join [Saint] [0] (~Saint]@unaffiliated/saint/x-8516940)
08:39:05 Nick pixelma_ is now known as pixelma (pixelma@rockbox/staff/pixelma)
09:00
09:02:32 Join stoffel [0] (~quassel@pD9E438F9.dip.t-dialin.net)
09:17:55 Join pretty_function [0] (~sigBART@123.252.213.191)
09:18:13 Quit Totalled (Read error: Connection reset by peer)
09:18:19 Join Totalled [0] (~Totalled@c-98-245-9-211.hsd1.co.comcast.net)
09:42:11 Join ender` [0] (krneki@foo.eternallybored.org)
09:42:33***Saving seen data "./dancer.seen"
09:45:26 Quit [Saint] (Remote host closed the connection)
09:57:38 Join megal0maniac [0] (~megal0man@196.213.53.210)
10:00
10:04:55 Join mgottschlag [0] (~quassel@HSI-KBW-46-223-232-253.hsi.kabel-badenwuerttemberg.de)
10:04:56 Quit mgottschlag (Changing host)
10:04:56 Join mgottschlag [0] (~quassel@reactos/tester/phoenix64)
10:28:01 Quit mystica555 (Remote host closed the connection)
10:30:08 Join mystica555 [0] (~Mike@71-211-209-19.hlrn.qwest.net)
10:36:20 Join [Saint] [0] (~sinner@unaffiliated/saint/x-8516940)
10:48:18 Quit pretty_function (Ping timeout: 246 seconds)
10:54:21 Quit mikroflops (Ping timeout: 246 seconds)
10:56:26 Join mikroflops [0] (~yogurt@h-34-21.a238.priv.bahnhof.se)
11:00
11:02:47 Join bertrik [0] (~bertrik@ip117-49-211-87.adsl2.static.versatel.nl)
11:02:47 Quit bertrik (Changing host)
11:02:47 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
11:04:38 Join pretty_function [0] (~sigBART@123.252.213.191)
11:12:49freqmodbertrik: hmm
11:13:04freqmodthe seeking does probably seek to the first packet which is the header
11:13:27 Quit fs-bluebot (Ping timeout: 245 seconds)
11:14:44 Join fs-bluebot [0] (~fs-bluebo@g231120239.adsl.alicedsl.de)
11:15:06 Quit bluebrother (Ping timeout: 245 seconds)
11:17:05 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
11:19:59 Join pamaury [0] (~quassel@cez63-2-88-164-98-172.fbx.proxad.net)
11:20:00 Quit pamaury (Changing host)
11:20:00 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
11:42:35***Saving seen data "./dancer.seen"
11:44:54 Join lebellium [0] (~chatzilla@i02m-212-194-176-149.d4.club-internet.fr)
11:49:18 Quit Wardo (Ping timeout: 246 seconds)
11:59:27 Join stoffel_ [0] (~quassel@pD9E4307D.dip.t-dialin.net)
11:59:42 Quit stoffel (Ping timeout: 246 seconds)
12:00
12:29:08 Join Ward [0] (~Mirandaha@bpb01-1-88-162-4-186.fbx.proxad.net)
12:29:33 Nick Ward is now known as Guest49978 (~Mirandaha@bpb01-1-88-162-4-186.fbx.proxad.net)
12:31:35 Quit pretty_function (Read error: Connection reset by peer)
12:31:45 Join pretty_function [0] (~sigBART@123.252.213.191)
12:34:05 Join n1s [0] (~n1s@nl118-168-30.student.uu.se)
12:34:06 Quit n1s (Changing host)
12:34:06 Join n1s [0] (~n1s@rockbox/developer/n1s)
12:49:16 Quit pretty_function (Ping timeout: 264 seconds)
13:00
13:03:09bertrikplease ban user "loving" from the forums, posting spam
13:05:38megal0maniacThat was quick
13:07:09*n1s wonders why the opus codec doesn't just statically alloc its buffers, they are statically sized and live as long as the codec runs afaict
13:07:21n1sand malloc w/o free looks weird
13:07:42n1s(and the codec_free function doesn't actually do anything :))
13:09:05n1sand wow, it's a lot of files
13:10:43 Join y4n [0] (~y4n@unaffiliated/y4ndexx)
13:11:43n1sbertrik: are silk and celt separate codecs?
13:12:24bertrikno,
13:13:27bertriksilk is based on some skype codec, but it's heavily modified for use in opus. celt was a codec research project from xiph as I understand.
13:14:46bertrikroughly, silk does low-frequency linear prediction coding and celt does mdct stuff for the higher frequencies, opus is the combination of the two
13:14:57n1sah
13:15:21n1sany idea if it can use the mdct lib we have in the codec lib?
13:16:22bertrikI don't know, I vaguely remember either saratoga or jhMikeS discussing this with the opus guys, the mdct size used was different IIRC
13:21:00n1sseems the fixed point representation it uses is different too but not sure whow hard that would be to change
13:22:18n1smight be fun to write some asm for it but i guess i'll wait untill it's in the repo
13:42:37***Saving seen data "./dancer.seen"
13:53:27 Quit mgottschlag (Ping timeout: 246 seconds)
14:00
14:02:40 Quit stoffel_ (Ping timeout: 244 seconds)
14:13:18gevaertsbertrik: done
14:14:15gevaertsn1s: multithreading. We don't run a codec multiple times from the same process, but some people do. You can;t statically allocate buffers then
14:17:50n1sgevaerts: we do that in pretty much all codecs, i only saw mallocs in the rockbox specific main codec file, are there other places too?
14:18:43 Join petur [0] (~petur@rockbox/developer/petur)
14:21:10gevaertsAh, right. Yes, I think part of that will need static allocation, if only to get the right bits into iram
14:22:08n1sah yes, avoiding the horrible iram malloc thing used in vorbsi would be nice :)
14:22:49n1si guess i'll take a closer look at it when it gets pushed
14:36:55 Join Thra11 [0] (~thrall@23.123.112.87.dyn.plus.net)
14:39:21 Quit mikroflops (Ping timeout: 244 seconds)
14:43:46 Join mikroflops [0] (~yogurt@h-34-21.a238.priv.bahnhof.se)
14:44:24bertriknot everything can be static when decoding opus
14:44:26 Join stoffel [0] (~quassel@pD9E4307D.dip.t-dialin.net)
14:44:41bertriklike we don't always know how big some buffers need to be in advance
14:45:12bertrikyou might not like the mallocs, but it's very practical and doesn't require all kinds of deep changes the make later upstream sync a lot harder
14:45:46n1sso there are allocs in the actual codec? i was under the impression there wasn't
14:45:49bertrikthere is no malloc / free during the playback phase, only during init
14:46:12bertrikthere is an ogg layer too
14:46:31bertrikand the opus does a few mallocs, yes
14:46:33 Join pretty_function [0] (~sigBART@123.252.212.79)
14:47:05n1sah yes, managed to not think about the god damned ogg container
14:47:54bertrikI wonder how heavy the ogg stuff is, it does quite a bit of memcpy/memmoving to assemble pages
14:48:02bertrikand it does CRC checks
14:48:48n1si benchmarked disabling the crc, was about 1-2% of total cpu on cf for vorbis iirc
14:49:17bertrikok, interesting, I would have guessed more impact
14:50:39 Quit [Saint] (Remote host closed the connection)
14:51:46n1ssorry, about 1% on PP apparentyl
14:51:53n1shttp://www.rockbox.org/mail/archive/rockbox-dev-archive-2010-06/0303.shtml if you're interested
14:51:56bertrikWe don't often sync codecs with upstream, do we?
14:52:02n1snot really
14:53:12n1sapparently people want to keep the crc check for vorbis, some other formats have similar checksums but their decoders don't check it (flac iirc)
14:54:07n1smy feeling is that if you get files spontaneously corrupted, you have larger problems than playing back some garbage or crashing the codec
14:54:32 Join [Saint] [0] (~sinner@unaffiliated/saint/x-8516940)
14:56:10bertrikI agree
14:57:10n1sbut it's not very expensive (and the crc in tremor isn't exactly optimized, reading a char-at-a-time)
14:59:15 Quit mikroflops (Ping timeout: 246 seconds)
15:00
15:13:35 Quit megal0maniac (Quit: megal0maniac)
15:16:31 Join mikroflops [0] (~yogurt@h-34-21.a238.priv.bahnhof.se)
15:17:28 Quit n1s (Ping timeout: 264 seconds)
15:23:17 Join Thra11_ [0] (~thrall@31.185.149.61)
15:23:47 Quit Thra11 (Ping timeout: 244 seconds)
15:39:03 Join Thra11 [0] (~thrall@31.185.149.61)
15:39:55 Quit Thra11_ (Ping timeout: 246 seconds)
15:42:38***Saving seen data "./dancer.seen"
15:48:25 Join mgottschlag [0] (~quassel@HSI-KBW-46-223-232-253.hsi.kabel-badenwuerttemberg.de)
15:48:25 Quit mgottschlag (Changing host)
15:48:25 Join mgottschlag [0] (~quassel@reactos/tester/phoenix64)
16:00
16:00:21 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
16:05:37 Join XavierGr [0] (~xavier@rockbox/staff/XavierGr)
16:23:35 Quit pretty_function (Ping timeout: 240 seconds)
16:38:18 Quit mc2739 (Ping timeout: 260 seconds)
16:44:43 Join pretty_function [0] (~sigBART@123.252.212.79)
16:46:45 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
16:50:21 Quit Guest49978 (Quit: Blarglarg)
16:54:45 Join n1s [0] (~n1s@rockbox/developer/n1s)
16:55:09 Quit benedikt93 (Quit: Bye ;))
17:00
17:01:01 Quit petur (Quit: Leaving)
17:03:29 Join Rower85 [0] (husvagn@v-413-alfarv-90.bitnet.nu)
17:14:18 Quit [Saint] (Read error: Connection reset by peer)
17:15:39 Join [Saint] [0] (~sinner@unaffiliated/saint/x-8516940)
17:18:43[Saint]Hmmmm...
17:24:13[Saint]the yp-r0 toolchain is pretty nuts.
17:42:42***Saving seen data "./dancer.seen"
17:42:46 Quit pretty_function (Ping timeout: 244 seconds)
17:44:51 Quit n1s (Quit: Ex-Chat)
17:53:22 Join perrikwp_ [0] (~quassel@cpe-024-163-024-033.triad.res.rr.com)
17:56:00 Quit perrikwp (Ping timeout: 246 seconds)
17:57:27 Join eckoit [0] (~ryan@50.65.10.24)
17:58:10 Quit stoffel (Read error: Operation timed out)
18:00
18:50:06 Quit eckoit (Quit: eckoit)
18:54:37 Join Thra11_ [0] (~thrall@75.56.113.87.dyn.plus.net)
18:56:28 Quit Thra11 (Ping timeout: 264 seconds)
19:00
19:11:17 Join eckoit [0] (~ryan@50.65.10.24)
19:16:51 Join pretty_function [0] (~sigBART@123.252.212.79)
19:22:50 Quit pretty_function (Remote host closed the connection)
19:42:44***Saving seen data "./dancer.seen"
19:50:41 Join stoffel [0] (~quassel@pD9E4307D.dip.t-dialin.net)
19:57:01 Join perrikwp [0] (~quassel@cpe-024-163-024-033.triad.res.rr.com)
19:59:11 Join perrikwp__ [0] (~quassel@cpe-024-163-024-033.triad.res.rr.com)
19:59:35 Quit perrikwp_ (Ping timeout: 240 seconds)
20:00
20:01:45 Quit perrikwp (Ping timeout: 244 seconds)
20:07:03 Nick perrikwp__ is now known as perrikwp (~quassel@cpe-024-163-024-033.triad.res.rr.com)
20:50:44 Quit Totalled (Quit: PETTAN PETTAN, TSURUPETTAN!)
20:52:23 Quit stoffel (Ping timeout: 244 seconds)
21:00
21:03:26 Join Wardo [0] (~Mirandaha@bpb01-1-88-162-4-186.fbx.proxad.net)
21:03:40 Join pretty_function [0] (~sigBART@123.252.212.79)
21:04:17 Join lebellium_ [0] (~chatzilla@i02m-212-194-176-149.d4.club-internet.fr)
21:05:18 Join lebellium__ [0] (~chatzilla@i02m-212-194-176-149.d4.club-internet.fr)
21:06:30 Join lebellium___ [0] (~chatzilla@i02m-212-194-176-149.d4.club-internet.fr)
21:06:30***Alert Mode level 1
21:06:30DBUGEnqueued KICK lebellium
21:06:30DBUGEnqueued KICK lebellium_
21:06:30***Alert Mode level 2
21:06:30DBUGEnqueued KICK lebellium__
21:06:30DBUGEnqueued KICK lebellium___
21:06:30***Alert Mode level 3
21:07:05 Quit lebellium (Ping timeout: 246 seconds)
21:07:10 Nick lebellium___ is now known as lebellium (~chatzilla@i02m-212-194-176-149.d4.club-internet.fr)
21:07:10DBUGEnqueued KICK lebellium
21:07:10***Alert Mode level 4
21:08:59 Quit lebellium_ (Ping timeout: 250 seconds)
21:09:51 Quit lebellium__ (Ping timeout: 250 seconds)
21:10:37 Quit Thra11_ (Ping timeout: 272 seconds)
21:17:11***Alert Mode OFF
21:40:50 Join perrikwp_ [0] (~quassel@cpe-024-163-024-033.triad.res.rr.com)
21:42:48***Saving seen data "./dancer.seen"
21:43:31 Quit perrikwp (Ping timeout: 246 seconds)
22:00
22:30:22 Quit pretty_function (Read error: Operation timed out)
22:36:59 Join TheDarkPirate [0] (~ricardo@190.202.154.188)
22:38:10 Quit TheDarkPirate (Client Quit)
22:38:18 Join TheDarkPirate [0] (~ricardo@190.202.154.188)
22:44:34 Quit y4n (Quit: PANTS OFF!)
22:48:12 Quit mystica555 (Remote host closed the connection)
23:00
23:11:41 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
23:27:14 Quit perrikwp_ (Read error: Connection reset by peer)
23:39:11 Quit benedikt93 (Quit: Bye ;))
23:42:51***Saving seen data "./dancer.seen"
23:53:48 Quit jfc (Ping timeout: 248 seconds)

Previous day | Next day