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 2012-10-07

00:12:23 Join bertrik [0] (~quassel@rockbox/developer/bertrik)
00:14:30 Quit guymann (Quit: Lost terminal)
00:14:56 Join guymann [0] (~c@unaffiliated/guymann)
00:32:21n1shmm, forgot to put numbers in last push, it saves about 1.6MHz on cf and 0.5 on pp5024
00:33:24 Quit lebellium (Quit: ChatZilla 0.9.89 [Firefox 16.0/20121002073616])
00:33:52bertrikwhat are the absolute numbers now?
00:34:43n1sfor the 64kbps file: cf 153.03% (81.15MHz), pp5024 117.75%(67.94MHz), amsv1 295.54% (83.91MHz)
00:39:09n1sBuschel's big patch has a few more small things that will probably save another MHz or two but nothing radical
00:39:40n1sthe next step will probably be writing some of the fft stuff in asm
00:50:02 Quit ender` (Quit: Connection Reset by Gypsies with Wire Cutters)
01:03:20derfWhat are your targets?
01:04:15n1sI'll probably focus on the cf and pp5024 which is arm7tdmi first, since they are the slowest
01:04:42derfI meant, the speed targets.
01:04:56n1sas low as it goes? :)
01:05:14n1snah but more speed == better battery time
01:07:06n1sthe rule of thumb is that 120% realtime is enough to give some cycles to the buffering, pcm and ui to make it fairly usable but now we also need to resample, also people might want to use some dsp on it
01:07:57n1sprimary target is probably to have all reasonable bitrates working for both targts
01:08:18gevaertsderf: the good news is that it's already a lot better than ape at higher compresseion levels :)
01:08:25n1sboth targts == both players
01:08:47n1sheh, yeah ape is just crazy
01:22:23n1shmm, switching to noninterleaved 32 bit stereo should help cf a bit too
01:30:47 Quit n1s (Quit: Ex-Chat)
01:32:13soapis there a reasonable expectation that it will ever be as efficient as the "common" lossy files or are there factors which will always keep it significantly more resource intensive?
01:45:36 Quit bertrik (Remote host closed the connection)
01:55:53***Saving seen data "./dancer.seen"
01:58:15 Quit pamaury (Ping timeout: 264 seconds)
02:05:49 Quit perrikwp (Read error: Connection reset by peer)
02:12:15 Quit mgottschlag (Ping timeout: 246 seconds)
02:24:41 Nick Jack87 is now known as Jack87|Away (Jack87@nasadmin/admin/jack87)
02:25:17 Nick Jack87|Away is now known as Jack87 (Jack87@nasadmin/admin/jack87)
02:41:17 Quit Rower85 (Quit: Hmmm...)
02:45:16 Join mikroflops_ [0] (
02:48:37 Quit mikroflops (Ping timeout: 246 seconds)
03:08:33 Join Guest3539 [0] (~quassel@
03:10:22 Nick Guest3539 is now known as [Saint] (~quassel@
03:10:51 Nick [Saint] is now known as Guest71524 (~quassel@
03:13:19 Nick Guest71524 is now known as [Saint] (~quassel@
03:13:49 Nick [Saint] is now known as Guest92404 (~quassel@
03:13:58 Quit Guest92404 (Changing host)
03:13:58 Join Guest92404 [0] (~quassel@rockbox/user/saint)
03:14:22 Nick Guest92404 is now known as [Saint] (~quassel@rockbox/user/saint)
03:14:39[Saint]that was annoying.
03:19:00 Part [Saint] ("Part")
03:20:57 Join [Saint] [0] (~quassel@rockbox/user/saint)
03:50:49 Join ParkerR_ [0] (znc@unaffiliated/parkerr)
03:55:56***Saving seen data "./dancer.seen"
04:07:24 Quit pixelma (Disconnected by services)
04:07:25 Join pixelma_ [0] (pixelma@rockbox/staff/pixelma)
04:07:25 Join amiconn_ [0] (amiconn@rockbox/developer/amiconn)
04:07:25 Quit amiconn (Disconnected by services)
04:07:27 Nick pixelma_ is now known as pixelma (pixelma@rockbox/staff/pixelma)
04:07:30 Nick amiconn_ is now known as amiconn (amiconn@rockbox/developer/amiconn)
04:27:18 Join icarusfactor [0] (~factor@
04:30:05 Quit factor (Ping timeout: 264 seconds)
04:32:30 Join sidney_ [0] (
04:35:44sidney_I have an ipod classic 30 GB Model No A1136 will rockbox work on this ipod?
04:36:19[Saint]There's no such thing. You have a "Classic iPod" (likely a Video) that is 30GB
04:36:26[Saint]and, if that is the case, yes.
04:37:22[Saint]Yes, it's a Video (not a Classic, but a "Classic iPod" {as in, old})
04:37:36sidney_saint thanks and is there a feature set page
04:37:48sidney_for rockbox i mean
04:38:26[Saint]You'll likely want the manual.
04:38:46[Saint]The features can vary wildly between supported devices.
04:38:47 Join mikroflops [0] (
04:39:35 Quit TheSphinX^ (Read error: Operation timed out)
04:42:28 Quit mikroflops_ (Ping timeout: 248 seconds)
04:46:13 Quit dfkt (Quit: -= SysReset 2.55=- Sic gorgiamus allos subjectatos nunc.)
04:46:23 Join TheSphinX^ [0] (~briehl@
04:48:18 Quit Epicanis (Quit: must sleep)
04:57:25 Join mikroflops_ [0] (
04:58:50 Quit mikroflops (Read error: Operation timed out)
05:04:45 Quit Totalled (Remote host closed the connection)
05:05:05 Join Totalled [0] (
05:05:22sidney_Saint thank you. Think I'll go to windows and try it
05:06:19 Quit sidney_ (Quit: Leaving)
05:07:57[Saint]Ok, if it floats your boat.
05:15:02 Quit [7] (Disconnected by services)
05:15:10 Join TheSeven [0] (~quassel@rockbox/developer/TheSeven)
05:35:34 Nick icarusfactor is now known as factor (~factor@
05:55:59***Saving seen data "./dancer.seen"
06:09:42 Quit Horscht (Quit: Verlassend)
06:18:52ParkerR_Been using Rockbox on my old Nano 1g. Loving it.
06:45:30 Quit prof_wolfff (Read error: Operation timed out)
06:53:41 Join JdGord [0] (~AndChat80@
06:56:12 Quit JdGord (Read error: Connection reset by peer)
07:00:40 Join JdGord [0] (~AndChat80@
07:01:48 Quit JdGord (Client Quit)
07:07:26 Join kevku [0] (
07:07:37 Quit XavierGr (Ping timeout: 245 seconds)
07:11:11 Quit TheSeven (Disconnected by services)
07:11:21 Join [7] [0] (~quassel@rockbox/developer/TheSeven)
07:44:52 Quit factor (Ping timeout: 248 seconds)
07:49:27 Join factor [0] (~factor@
07:56:01***Saving seen data "./dancer.seen"
08:01:23 Join pretty_function [0] (~sigBART@
08:40:02 Quit bootinfdsds (Ping timeout: 246 seconds)
08:41:41 Join bootlfdhfrsa [0] (~Prmhfhfx@
08:46:15 Quit bootlfdhfrsa (Ping timeout: 260 seconds)
08:47:20 Join bootlfdhfrsa [0] (~Prmhfhfx@
08:52:31 Quit n17ikh (Ping timeout: 255 seconds)
08:59:10 Join n17ikh [0] (~n17ikh@
08:59:10 Quit n17ikh (Changing host)
08:59:10 Join n17ikh [0] (~n17ikh@unaffiliated/n17ikh)
09:25:41 Join n1s [0] (~n1s@rockbox/developer/n1s)
09:25:51 Join stoffel [0] (
09:40:16 Join wodz [0] (
09:47:52 Join bertrik [0] (~quassel@rockbox/developer/bertrik)
09:49:38 Quit n1s (Read error: Connection timed out)
09:56:02***Saving seen data "./dancer.seen"
10:04:56 Join Buschel [0] (
10:08:34Buscheln1s: one small patch can be applied independent of the upcoming asm'ing -> loops are exchanged with calls of OPUS_COPY. this saves ~0.8 MHz on pp
10:16:16 Join Rower85 [0] (husvagn@
10:26:07 Quit Buschel (Ping timeout: 255 seconds)
10:33:47wodzI have proof-of-concept veneer insertion code - time to test on target :-)
10:34:10 Join lebellium [0] (
10:35:04 Join ender` [0] (
10:37:23 Join Buschel [0] (
10:48:52 Quit pretty_function (Ping timeout: 248 seconds)
10:50:55[Saint]My mother in law just had veneer blinds put could've saved her a fortune ;)
10:57:19wodz[Saint]: if veneer in form of ldr pc, [pc, #-4] .word TARGET_ADDRESS will work for her...
11:00:34 Join pretty_function [0] (~sigBART@
11:01:37 Join fyre^OS [0] (
11:04:17 Quit fyrestorm (Ping timeout: 264 seconds)
11:06:47 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
11:08:37 Quit TheSphinX^ (Ping timeout: 252 seconds)
11:09:57 Join TheSphinX^ [0] (
11:14:58 Quit pretty_function (Remote host closed the connection)
11:45:49 Nick scorche` is now known as scorche (~scorche@rockbox/administrator/scorche)
11:48:35 Quit bluebrother (Disconnected by services)
11:48:40 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
11:50:23 Quit fs-bluebot (Ping timeout: 260 seconds)
11:50:56 Join n1s [0] (
11:50:57 Quit n1s (Changing host)
11:50:57 Join n1s [0] (~n1s@rockbox/developer/n1s)
11:51:48 Join fs-bluebot [0] (
11:56:05***Saving seen data "./dancer.seen"
12:08:37 Quit pamaury (Ping timeout: 246 seconds)
12:11:33wodzha - doom now works
12:11:37n1sBuschel: yeah, i was planning to do that last bit. Are you working on anything else?
12:11:51wodzso everything else is irrelevant :P
12:12:06Buscheln1s: i am taking a look at the MULC
12:12:14BuschelCMUL_C macros
12:13:00Buschelit ssems we can adapt some easy asm stuff from codeclib. they look very similar to PROD31
12:13:26[Saint]wodz: congrats!
12:13:48[Saint]As we all know, DOOM is the only thing that matters. Well done!
12:13:49 Join prof_wolfff [0] (
12:14:01 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
12:14:43n1sBuschel: yeah, there are some cmul functions somewhere in codeclib too
12:15:11Buscheln1s: they all use >>31 instead of >>15
12:45:08bluebrotherwodz: another reason not to remove doom then? :(
12:46:26n1si'm thinking of changing the output format to noninterleaved 32 bit, it will save the dsp from converting to noninterleaved 32 bit and also the redundant clipping done in opus as the dsp clips at the end of processing anyway
12:47:16bluebrotherbtw, how fast is opus these days? I tried playing a file yesterday which mostly worked but the UI was totally unresponsive (on mini2g that is)
12:47:25bluebrotherI guess that is still to be expected?
12:47:57n1sshould also make it easier to write an efficient asm version of deemphasis
12:49:23n1sbluebrother: there are numbers in today's log at 00:34
12:49:49*bluebrother scrolls back
12:50:45bluebrothernice, thanks
12:53:59[Saint]I didn't have any problem with 64k opus on RaaA that I noticed, but...that's kinda meaningless I guess.
12:54:10 Quit [7] (Disconnected by services)
12:54:17[Saint]a lot more RAM and CPU than most targets.
12:54:18 Join TheSeven [0] (~quassel@rockbox/developer/TheSeven)
12:54:29wodzIf anyone wants to help with g#326 please test on ARM other then PP5020 (but not thumb build) or on CF and report the results.
12:54:32fs-bluebotGerrit review #326 at : ELF loader for plugins by Marcin Bukat (changes/26/326/4)
12:54:42Buscheln1s: whar will you be looking at?
12:55:31n1sBuschel: i'm not sure, i'd like to change the output format as i said above but it looks rather complicated
12:59:19n1smaybe asm of the comb_filter instead
13:01:28Buschelhmm, if I understand right (from a qucik look) changing the output format would make deempahsis easier and maybe faster. would'nt simply y+=C migrate to y+=1?
13:02:24n1syes, and the clipping done by SIG2WORD16 would be unnecessary
13:02:50Buscheland multiple load/str :)
13:03:15n1sand the dsp wouldn't need to convert the samples
13:04:20Buschelso, where is the complicated part?
13:06:30Buschelin fact this sounds more interesting that asm'ing comb_filter. deemphasis cosnumes 2x as much.
13:07:26n1syeah, ok maybe not very complicated but it's passing the buffer pointer around a lot and there's the silk side of it too
13:09:36Buschelif it is easy to do for celt, this is at least a good starting point. in worstr case we could just add a converter for silk
13:09:52n1sok, i'll give it a shot
13:13:20lebelliumis there a specific reason for which only very old Archos models are supported or is it just a lack of interest for newer models?
13:14:04n1snoone ported rb to newer archos players
13:14:43n1ssame reason as for any other device we don't support
13:17:18lebelliumOk. I asked that because I was surprised to see that the old 2001 Archos Jukebox Multimedia is not supported while all models that have been released just before are supported. So I thought maybe it's just because the hw/sw is completely different
13:17:43n1sthe hardware at least is very different
13:17:53n1siirc that had an arm with some dsp
13:19:11lebelliumhum ok
13:19:32n1sthe archos models we support have sh cpu's and dedicated hardware mp3 decoders
13:19:59[Saint]I think interest plays a large part.
13:20:13[Saint]Someone has to do the work.
13:20:19n1syes, definitely
13:20:21lebelliumso the same basic work could have be reused for all supported Archos players ?
13:20:48n1slebellium: yes, the supported archos players are similar in hw
13:23:42lebelliumok and a more general question. Why are there so many pre-development ports but only a few led to unstable port. Is it the work too difficult? I'm not a dev at all so it's hard for me to understand those things. Why people start a port but give up so quickly? :S
13:24:35wodzlebellium: It is *VERY* hard work and real life has its right unfortunately
13:24:48n1svarious reasons, but usually a lack of interest from others and limited free time and fading interest i guess
13:25:43n1ssome ports also come up against stumbling blocks like complicated FTL
13:26:22lebelliumare there many stable/unstable ports due to a single guy or are most of working ports due to team work?
13:26:51[Saint]essentially all are a team work.
13:26:56n1smost of the stable ones are teamwork i'd say
13:27:35[Saint]there's no one device where a single dev had sole input.
13:27:37n1ssome ports were made by single people but then usually on hardware that was very similar to other ports
13:27:47[Saint]such is the nature of Rockbox.
13:28:12[Saint]reuse code when you can...'cos, why not...right?
13:28:12n1sexcept perhaps wodz' mpio port :)
13:28:36[Saint]Ah, yes...he did his own theme there too :)
13:28:41lebelliumok so basically if there is a lack of interest from others, a port can't be led to unstable/stable
13:28:59wodzn1s: It was heavily based on iriver/iaudio work
13:30:00wodzlebellium: It can but the timeframe will probably make it uninteresting for others
13:30:07AlexPlebellium: It needs someone who both knows what they are doing and wants to do it
13:30:31n1swodz: yeah of course but still
13:30:32[Saint]...with ~20 hours a week spare.
13:30:45AlexPIf it gets to a certain point (usually the low level hardware is "done"), then it will benefit from the higher levels of Rockbox same as anything else
13:30:49wodzpdbox makefile is just stupid
13:31:00wodzand incorrect at the same time :-)
13:31:17 Join nuskool [0] (
13:31:55[Saint]then mend it dear Liza. dear Liza, dear Liza, mend it. ;)
13:32:19lebelliumI'm wondering: as the NW-A808 is Linux-based like the YP-R0, wouldn't be quite easy to port RaaA?
13:32:32AlexPlebellium: It isn't quite right, but you could think of it a bit like drivers for the hardware and then the OS e.g. Windows on top
13:32:45AlexPlebellium: Maybe :)
13:32:58AlexPmaybe not
13:33:27AlexPWhen you are talking about Rockbox being hosted and not being the firmware in its own right, then that is a different kettle of fish
13:33:56wodzlebellium: it there is a way to prepare upgrade image it should be simpler then fully fledged native port.
13:34:23[Saint]if it followed the YP-R0, it would be the sole firmware. But, ...yeah.
13:34:37[Saint]there's likely little similarity between them.
13:35:01lebelliumI think Sony players have great hardware with tactile buttons, no stupid touchpad or things like that, so it's a pity there is no supported Sony player
13:35:24 Quit ChanServ (*.net *.split)
13:35:33[Saint]Buy one and do something about it :)
13:36:55lebelliumahaha, giving my coding level (limited to the skins engine language), it would be better to convince a dev that he should port rockbox on a Sony player. That"s why I try to find an easy target, like a linux-based for instance :D
13:37:35AlexPI don't think you'll persuade anyone to do the port for you unless they want to anyway.
13:37:51AlexPYou could of course persuade them what a nice player it is so they might want to :)
13:37:56[Saint]Yeah...not really. One is likely to happen (you doing something), the other...not so much.
13:38:15AlexPBut it'd help if you did a bit of research to show that it was at least likely to be possible first
13:38:35[Saint]A dev will work on something that interests them, its pretty hard to say "port X for me" and have it happen.
13:38:46[Saint]This isn't Android, lol :)
13:39:33lebelliumI don't mean convince a dev to port for me, I mean convince a dev that Sony players have great hw but poor sw and therefore need a RB port :p
13:40:03AlexPSemantics are fun :)
13:40:15[Saint]That's all well and good, but it still doesn't mean anyone will work on it.
13:40:28lebelliumfor sure
13:40:33lebelliumwas juste like ... a dream
13:40:39[Saint]lots of devices have nice hardware, crap software, ...and no Rockbox :)
13:40:46lebelliumI know people here are more sansa-oriented
13:40:51AlexPNot really
13:41:00AlexPIt is just sansa make a range of decent cheap hardware
13:41:12AlexPSo they happen to have got a lot of interested people recently
13:41:17[Saint]People here are <stuff they are interested in> oriented.
13:42:45lebelliumbut is the work on sansa players easier because it's always similar hw/sw?
13:42:56AlexPAs it isn't
13:44:52 Join ChanServ [0] (ChanServ@services.)
13:44:52Mode"#rockbox +o ChanServ " by
13:45:04lebelliumBertrik ported RB to the Zip so fast that it looked easy from a non-dev point of view :p
13:45:14AlexPSome might be similar
13:45:20AlexPI don't know about the specifics
13:45:26AlexPBut not all of them are
13:45:51AlexPThe quickest port was amiconn in a few hours :)
13:46:04AlexPAn iaudio IIRC
13:46:04[Saint]Hahaha, yeah.
13:51:12lebelliumsome APE files make the R0 crash
13:51:24lebelliumthe R0 likes crashing with rockbox :D
13:51:31AlexPI'd use a more sensible lossless format
13:51:38AlexPlike flac, or wavpack
13:51:50AlexPOne that doesn't need a supercomputer to decode it :)
13:51:55lebelliumI don't use it. I just tried the files used by a russian user
13:53:17lebelliumI assume APE is not supposed to make the R0 crash though
13:53:40lebelliumeven if it's not a supercomputer, it has a great 532MHz CPU
13:54:57amiconnAlexP: Quickest of my ports was Mini G2 iirc
13:55:28amiconnAbout 10 hours from receiving the deviceto rockbox running (usably)
13:56:06***Saving seen data "./dancer.seen"
14:00:03lebelliumthe APE files also make the Zip crash. Should I tell the russian user to open a bug tracker?
14:01:56amiconnThe mini g2 was rather easy because of the ipl project and its similarity to other, at that point already supported ipods
14:04:00AlexPlebellium: He probably doesn't need to go that far, just opening a bug report on the existing tracker should suffice :)
14:04:07AlexPBut yeah, if one doesn't exist
14:04:23AlexPVarious levels of ape may not play realtime, but it shouldn't crash Rockbox
14:04:49lebelliumok thx
14:12:02 Quit Provel (Ping timeout: 260 seconds)
14:12:46 Join mikroflops [0] (
14:13:25wodzcould someone explain me the reds?
14:14:39wodzhmm missing -lm probably
14:15:26gevaertsThat shouldn't need -lm
14:16:03 Quit mikroflops_ (Ping timeout: 244 seconds)
14:16:35gevaertswodz: are you sure $(PDBOX_OBJ)
14:16:39gevaertsis the same as *.o?
14:16:45 Quit stoffel (Remote host closed the connection)
14:17:54wodzits exactly the same as doom
14:18:14wodzit looks like it needs -lm in final link in sims
14:19:11gevaertsThe sims shouldn't ever need -lm
14:19:32wodzlook at line 26 of pdbox.make
14:20:24wodzthe errors are about unreferenced math functions
14:22:38wodzso how it works on target?
14:22:54gevaertshm, now that I think about it that was probably the main reason that special rule was there in the first place
14:23:13 Quit soap (Ping timeout: 260 seconds)
14:23:17gevaertsOn target it probably uses different code
14:24:07wodzits in pdbox.h but why on earth it doesn't use the same code on target and in sim when it can?
14:24:51 Quit bertrik (Remote host closed the connection)
14:25:16gevaertsNo idea. Feel free to fix it, but I personanlly want to stay as far away as I can from the pdbox build system...
14:26:17gevaertsActually the main question is why the sim falls under APP_TYPE I think
14:31:08wodzwill look at it after dinner
14:31:35gevaertsI suspect that CONFIG_PLATFORM & PLATFORM_HOSTED in pdbox.h should be something else
14:34:14wodzmy point is that it provides int implementations for various math functions, so I don't see any reason to depend on math.h on PC
14:35:08 Join soap [0] (~soap@rockbox/staff/soap)
14:36:22gevaertsFor app builds, the system-provided implementation may well be a lot faster
14:36:44 Join bertrik [0] (~quassel@rockbox/developer/bertrik)
14:37:05 Join mgottschlag [0] (~quassel@reactos/tester/phoenix64)
14:37:32 Join mikroflops_ [0] (
14:37:33wodzbut do we care about this in sim?
14:38:24 Quit guymann (Quit: aasdsf)
14:38:24gevaertsNo, the sim should be close tho target
14:40:55 Quit mikroflops (Ping timeout: 260 seconds)
14:41:58 Join guymann [0] (~c@unaffiliated/guymann)
14:47:35 Quit sciopat (Quit: Quitte)
14:51:29wodzgevaerts: it will always use built-in math functions, but is it really a problem?
14:53:19gevaertsI'd say talk to kugel or one of the other app people about this
14:54:11gevaertsHmm, is pdbox actually built for apps?
14:54:47wodzyou ask me?
14:55:42wodzandroid build doesn't have red so I'd say no
14:56:03gevaertsIt's build for YP-R0
14:56:47 Join sciopath [0] (
14:56:52wodzI propose to either commit quick fix from pastie and then talk with kugel or revert the commit
14:57:32wodzgevaerts: whats your opinion?
14:57:37gevaertsSounds reasonable. That pastie has C++-style comments though :)
14:59:29wodzI know, will fix.
15:00:49 Quit bertrik (Read error: Connection reset by peer)
15:01:57 Join bertrik [0] (~quassel@rockbox/developer/bertrik)
15:04:55*wodz crosses fingers
15:06:04wodzIs pdbox documented in manual? I can't find it.
15:07:55 Quit Buschel (Quit: ChatZilla [Firefox 15.0.1/20120905151427])
15:11:09wodzbtw. when testing plugins it is sooooo much frustrating that there is no consistent quit behavior.
15:20:33[Saint]There are some really bad (totally non-obvious) ones, yes.
15:55:43 Join mikroflops [0] (
15:56:09***Saving seen data "./dancer.seen"
15:59:22 Quit mikroflops_ (Ping timeout: 256 seconds)
16:04:43n1swell, switching to using 32 bit interleaved samples and skipping the clipping and scaling saves over 2MHz on cf but i can't get noninterleaved to work :/
16:09:38n1sturns out it is rather complicated as the codec does some processing on the celt output and naturally everything expects the interleaved format
16:20:17 Join pretty_function [0] (~sigBART@
16:39:41n1sderf: our dsp layer uses noninterleaved 32 bit samples so if a codec outputs something else it converts them. The dsp then scales down to 16 bit and does clipping at the end of processing
16:40:13 Join mortalis [0] (~mortalis@
16:40:54derfAh, okay.
16:41:27derfSo these are not Q31 (i.e., you allow magnitudes greater than 1.0)?
16:42:39 Quit GeekShadow (Changing host)
16:42:39 Join GeekShadow [0] (~antoine@reactos/tester/GeekShadow)
16:42:53n1sno they are q31
16:44:24 Quit bertrik (Remote host closed the connection)
16:44:59derfOkay, then I guess I don't understand how you can just do clipping at the end.
16:46:21n1sthinking about it i guess i can't
16:46:38n1sstill should save the dsp sample conversion
16:49:17 Join bertrik [0] (~quassel@rockbox/developer/bertrik)
16:50:14n1shmm,i might be confused
16:52:00n1sappears to be s4.27 actually
16:53:06 Quit pamaury (Ping timeout: 240 seconds)
16:53:35derfOkay, that makes more sense.
16:55:04derfBut anyway, yeah, if you want non-interleaved output, you'll have to rewrite a bunch of stuff in opus_decoder.c (in addition to all the changes in celt.c).
16:55:51n1syeah i saw that
17:08:07 Quit factor (Read error: Connection reset by peer)
17:15:09 Join factor [0] (~factor@
17:18:16n1sderf: btw the clipping could be made faster by doing if(x != (int16_t)x) x = 0x7fff ^ (x>>31);
17:22:06 Quit factor (Read error: Connection reset by peer)
17:22:13 Join factor [0] (~factor@
17:23:07derfDepends. On architectures where you actually have min and max instructions, the current code is 2 instructions with no branches.
17:26:33derfEven on ARM, I think yours is only fewer if you manage to get 0x7FFF in a register.
17:26:34 Quit zahari_bgr (Read error: Connection reset by peer)
17:33:57 Join zahari_bgr [0] (~zahari.yu@
17:44:46 Join mikroflops_ [0] (
17:48:10 Quit mikroflops (Ping timeout: 256 seconds)
17:50:29 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
17:56:11***Saving seen data "./dancer.seen"
17:56:28 Quit pretty_function (Ping timeout: 246 seconds)
18:02:02 Quit zoktar (Ping timeout: 260 seconds)
18:17:03 Quit bertrik (Remote host closed the connection)
18:25:11 Join bertrik [0] (~quassel@rockbox/developer/bertrik)
18:27:04 Join XavierGr [0] (XavierGr@rockbox/staff/XavierGr)
18:40:36 Join zoktar [0] (
18:54:53 Join hype [0] (~hype@
19:03:46 Quit user890104 (Ping timeout: 256 seconds)
19:08:23 Join user890104 [0] (Venci@
19:10:44 Quit mortalis (Remote host closed the connection)
19:10:58 Quit n1s (Read error: Connection timed out)
19:17:29 Quit factor (Read error: Connection reset by peer)
19:19:40 Join pretty_function [0] (~sigBART@
19:34:14 Join factor [0] (~factor@
19:37:32 Join Epicanis [0] (
19:44:46 Join mikroflops [0] (
19:48:03 Quit mikroflops_ (Ping timeout: 248 seconds)
19:54:12 Quit mgottschlag (Read error: Connection reset by peer)
19:56:14***Saving seen data "./dancer.seen"
19:58:58 Quit user890104 (Quit: .)
20:00:01 Join user890104 [0] (Venci@
20:05:23 Join liar [0] (
20:10:02 Join lorenzo92 [0] (
20:18:27 Quit kadath (Ping timeout: 248 seconds)
20:33:00 Join n1s [0] (
20:33:00 Quit n1s (Changing host)
20:33:00 Join n1s [0] (~n1s@rockbox/developer/n1s)
20:34:12n1sderf: the condition should rarely be true so the test matters more than the actual clipping code
20:35:24n1son cf the min and max are both conditional branches i haven't checked what gcc does for arm
20:35:37derfProbably something terrible.
20:35:52derfBut it _should_ be two instructions each with no branches.
20:36:36n1swhat arches have min and max instructions?
20:36:57derfPretty much any DSP.
20:44:03derfBut I mean, if you're on ARMv5, just the test requires a shift, shift+cmp, and branch.
20:44:04 Join kadath [0] (
20:44:54derf(because there's no good way to convert to a 16-bit value)
20:45:36derfSo even if the branch is perfectly predicted, you're saving one instruction off the cmp, cmov, cmp, cmov branchless approach.
20:46:06derfAnd that assumes you never clip.
20:47:23 Join bootinfdsds [0] (~Prmhfhfx@
20:47:43derf(and of course, if you're on ARMv6 or better, what you really want is just SSAT)
20:47:47n1sright, on cf it's the reverse, a mov with 16->32 bit sign extension, a cmp and a conditional branch instead of two (more expensive) compares and two conditional branches
20:48:34n1sbut it's a rather obscure platform :)
20:50:04n1syeah the armv6 stuff is nice but we have only one such target and it's plenty fast
20:50:07 Quit lorenzo92 (Quit: ChatZilla 0.9.89 [Firefox 15.0.1/20120907231657])
20:50:30 Quit bootlfdhfrsa (Ping timeout: 260 seconds)
20:51:48n1sgetting it to output noninterleaved 32 bit samples would help everywhere though, even by just reducing memory reads and writes
20:54:13 Join LambdaCalculus37 [0] (~LambdaCal@rockbox/staff/LambdaCalculus37)
20:55:31 Join Keripo [0] (
20:59:08 Quit n1s (Quit: Ex-Chat)
21:30:55 Quit guymann (Quit: :ยป)
21:41:29 Quit liar (Ping timeout: 246 seconds)
21:53:10 Join baptistecl [0] (
21:56:16***Saving seen data "./dancer.seen"
22:08:22 Quit baptistecl (Quit: ChatZilla [Iceweasel 3.5.16/20120921165143])
22:10:17 Quit mc2739 (Ping timeout: 246 seconds)
22:12:22 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
22:21:00 Join eckoit [0] (
22:23:35 Join guymann [0] (~c@unaffiliated/guymann)
22:32:56 Quit pretty_function (Ping timeout: 246 seconds)
22:34:19 Join amayer [0] (
22:41:46 Join lorenzo92 [0] (
22:42:26lorenzo92is it possible to set the current (file)browser folder position? i.e. after removing the SD in ypr0 it should be needed to re-set root folder on the filemanager otherwise it gets stucked
22:42:33lorenzo92perhaps it's also a bug to fix
22:42:54 Join pamaury [0] (
22:42:55 Quit pamaury (Changing host)
22:42:55 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
22:43:19lorenzo92to explain better: I'm in mnt/mmc/folder1, playing a file. I remove the sd card in WPS. going back doesn't work, neither going into filebrowser from menu
22:43:27lorenzo92it works only if I reinsert microSD
22:45:35lorenzo92well seing the code, it's not a bug for HAVE_HOTSWAP targets, perhaps need to fix that for host targets too :)
22:45:39 Quit lorenzo92 (Client Quit)
22:47:46 Quit wodz (Quit: Leaving)
22:50:03 Quit XavierGr (Ping timeout: 252 seconds)
22:51:35 Join XavierGr [0] (XavierGr@rockbox/staff/XavierGr)
22:51:43 Quit Rower85 (Read error: Connection reset by peer)
22:58:25 Join mikroflops_ [0] (
23:00:56 Quit XavierGr (Ping timeout: 245 seconds)
23:01:33 Join XavierGr [0] (XavierGr@rockbox/staff/XavierGr)
23:01:48 Quit mikroflops (Ping timeout: 260 seconds)
23:08:01 Quit XavierGr (Ping timeout: 245 seconds)
23:08:17 Join XavierGr [0] (~xavier@rockbox/staff/XavierGr)
23:09:46 Quit zchs (Remote host closed the connection)
23:19:26 Quit benedikt93 (Quit: Bye ;))
23:30:46 Quit zahari_bgr (Read error: Connection reset by peer)
23:31:10 Join pamaury_ [0] (
23:31:43 Quit pamaury (Ping timeout: 246 seconds)
23:36:25 Quit ps-auxw (Ping timeout: 272 seconds)
23:36:43 Join zahari_bgr [0] (~zahari.yu@
23:36:45 Join ps-auxw [0] (~arneb@2001:470:c807:0:1532:4e5f:2ad3:4123)
23:53:37 Join mgottschlag [0] (~quassel@reactos/tester/phoenix64)
23:56:17***Saving seen data "./dancer.seen"

Previous day | Next day