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 2009-10-24

00:00:54kugelI also can't explain that. The only time I noticed a that slow UI on other targets was with caches disabled on my fuze, but you said they're definitely on
00:02:12 Join elinenbe [0] (
00:02:23*TheSeven experiencing PLL corruption
00:03:05TheSevenCLKCON and PLL0PMS seem to be garbled up sometimes during boosting
00:03:12kugelmaybe you can implement led() (so that the disk access icon appears in the statusbar) to exclude excessive storage accesses
00:03:17TheSevenmaybe we need to ramp Vcore a little? :-P
00:03:18 Join StealthyXIIGer [0] (
00:04:00 Join elinenbe__ [0] (
00:04:08 Quit elinenbe_ (Read error: 113 (No route to host))
00:04:09 Nick elinenbe__ is now known as elinenbe_ (
00:05:32JdGordon|TheSeven: slow ui scrolling could be slow lcd driver? or slow disk if it has to keep swapping out font glyphs?
00:06:18TheSevenJdGordon|: LCD is ~80-100FPS fullscreen RGB
00:06:26TheSevenquarter screen was around 300
00:06:39JdGordon|its not that then :)
00:06:51kugelthe font idea is good one, in fact that was another time I noticed slowish UI (I forget, some versions of the anti-aliased font patch caused huge slowdowns)
00:06:53TheSevenhow big is that glyph cache?
00:07:01TheSevenstorage is really a bottleneck on that one
00:07:04JdGordon|100K i tinhk
00:07:20JdGordon|shuold be plenty big enough if your filenes are all english
00:07:26JdGordon|dircache enabled?
00:07:37TheSevendircache doesn't seem to make any difference
00:07:53TheSeven(at least in the file browser)
00:08:03TheSevenbut the plugin menu behaved very weird with dircache
00:08:38 Quit bmbl ("Bye!")
00:16:16TheSevenin fact the dircache slowed scrolling down as soon as i scrolled to regions where i hadn't been before
00:16:23TheSevenwith dircache disabled, it's all fast
00:16:42TheSeven(well, as fast as the file browser at least)
00:17:05kugelI didn't have dirache on when I played around
00:17:45TheSevenwell, with dircache on it was like half a second for every scrolling step
00:18:10kugeluh, that's really bad
00:18:36kugelit seems accessing ram is slow
00:19:45TheSevenno, accessing flash
00:20:05kugelwell, you don't access flash if you have dircache on
00:20:07TheSevenor maybe non-builtin strings
00:20:25TheSevenall the builtin menus are just fine
00:20:47kugeldircache contains every entry (file and dirs), and is build during boot
00:20:58*kugel still suspects some cache problem
00:21:01 Join Sean [0] (
00:21:29TheSevenyes, i saw some splashf flash very briefly during boot
00:21:30 Nick Sean is now known as Guest24033 (
00:21:31TheSevenflash read is at ~1MB/s, which is quite bad, but sufficient
00:21:40 Join jdii [0] (
00:22:23TheSevenso it's non-builtin strings? how can that happen?
00:22:33kugelso unless you actually access the content of files, you never actually hit the storage
00:23:38kugeli really recommend implementing the disk access indicator
00:23:55TheSevenhow do i do that from the storage side?
00:24:35 Quit elinenbe (Read error: 113 (No route to host))
00:24:45TheSevenhmm there is a weird stub
00:24:51TheSevenvoid nand_led(bool onoff) { led(onoff); }
00:25:03pixelmaGuest24033: just ask your question (if it is Rockbox related)
00:25:08*TheSeven wonders what nand_led is good for then
00:26:06 Quit Utchybann ("I like core dumps")
00:26:26TheSevenhmm that thing is just unused crap. no idea where it comes from.
00:28:11 Join datty [0] (
00:28:21dattyrock that box
00:28:30dattyrock it!@
00:29:37Torneyeah we'll get right on that
00:30:14TheSevenkugel: ok, found something interesting
00:30:50TheSeventhe database top-level menu, where the submenus "Artist", "Album", "Genre", ... are, is also sluggish, but not other builtin menus like the main menu
00:30:54 Join nugged [0] (
00:31:41 Quit sbhsu (Read error: 110 (Connection timed out))
00:31:41TheSevenand that sluggishness is hardly related to the size of the list to scroll
00:33:40Guest24033I got one question about the v2 AMS series sansa, it says no usb support for it, does that mean i won't be able to sync music to it through usb or something else?
00:33:41dattyi would like to get into this dev stuff for my Fuze
00:33:49TheSevenneither does the length of the text nor the type of the line selector have any influence
00:33:58dattyI have some experience with AVR/PIC stuff, is it pretty similar with ARM?
00:34:28TheSevendatty: ARM is more like a desktop CPU compared to those microcontrollers, but of course, the basics are the same
00:34:37Tornedatty: a little. Almost all of rockbox is written in C, though, so the low level details of architecture are not too important for most things
00:34:52dattyi do prefer C over asm anyday
00:35:30Torneif you can program in C and are familiar with concepts like memory mapped devices that's more or less all you need
00:35:47TheSevenkugel: interesting... runtime-generated debug menu contents aren't sluggish either
00:35:53Torneactually knowing arm/m68k/coldfire is not important unless you end up in very specific parts of the code
00:35:54 Quit nugged ("CGI:IRC (Ping timeout)")
00:36:06 Quit Guest24033 ("CGI:IRC (EOF)")
00:36:07gevaertsGuest24033: it means you have to reboot to the original firmware for USB
00:36:40dattyi am used to AVR Studio and MPLAB, is there something similar with this?
00:37:15dattyI see "get the source, build, and change (patch)"
00:37:28TorneYes. You can use any editor/IDE you like..
00:37:56Tornebut a fully integrated system like AVR Studio is not really possible on platforms like DAPs where you have no code download or debug facilities
00:37:59dattyi guess i mean more for in circuit debugging, is this even possible?
00:38:03TheSevendatty: but of course the funny part is that there is no proper on-target debugging functionality :-)
00:38:10TorneNo, we basically have no on-target debugging
00:38:18TorneA few platforms have a JTAG port, if you rip them open
00:38:26Torneand solder stuff tot he right place.
00:38:27dattythen that makes things more difficult ;)
00:38:30Tornebut most don't.
00:38:39TheSevenwell, we have basic things like a log ring buffer, but nothing you could really call a debugger
00:38:59Tornedatty: there's various ways of logging debug messages, either to a buffer or over serial/usb, depending on hardware
00:39:06Tornebut that's basically all you get. no ICE/ICD
00:39:39TheSevenif we had ICD and proper datasheets, all those targets would be implemented in no time :-)
00:39:43dattyI'm interested in somehow interfacing directly with the ARM, has anyone tried that
00:39:50Tornedefine "interfacing directly"
00:39:56dattyTheSeven: yeah thats why I got interested !
00:39:58Torneyou mean with your mind, or what? :)
00:40:09dattyi was thinkin man whats takin so long :)
00:40:20dattybut yea its much harder
00:40:25dattyinterfacing directly with the pins
00:40:32Tornedatty: yeah, all our porting is done based on what little docs are published, and on reverse engineering the orignal firmware, and general guessing :)
00:40:36Tornedatty: er, to do what?
00:40:44TorneWe're already running code on the chip
00:40:49Torneyou can't really get more direct than that
00:41:14dattyyea but i'm sure that ARM has a program for in circuit debugging, no?
00:41:29dattymaybe my questions don't make sense, i've never worked on a device like this
00:41:32TorneARMs are not sepearate chips
00:41:34dattyso keep that in mind
00:41:42TorneThey are synthesized into the middle of a system-on-chip
00:41:46Tornewith peripherals and so on on the same die.
00:41:57TorneIf the JTAG pins are not brought out onto the SoC's pins there's nothing you can do
00:42:06TheSevendatty: ARM maybe has, but this doesn't mean it's accessible anywhere on the board or even SOC package, and sometimes those things are even fused out during production
00:42:07Torneor equally, if they are brought out but not connected to tracks
00:42:19Torneand yes, sometimes they get disabled in the factory with efuses
00:42:25dattyi see
00:42:40dattybut ARM922TDMI is a separate chip right
00:42:46TorneThere is no such thing as an ARM chip
00:42:56TorneThey haven't made physical ARm processors for over a decade
00:43:06TheSevenTorne: or that funny thing we found on the nano2g, that the data bus shuts down until reset with the first jtag clock edge to prevent you from dumping the bootrom :-)
00:43:26TheSeveni bet there is a secret way to bypass this, but only apple knows how.
00:43:43dattyo nano
00:43:56Tornedatty: All ARM processors on hardware made since the mid-90s are just a macrocell inside a larger ASIC
00:43:59TheSevenwell, probably apple, arm and samsung, in fact
00:44:07Tornedatty: look up System-on-Chip on wikipedia :)
00:44:14TorneTheSeven: ARM are unlkely to know :)
00:44:28Tornethey stay out of SoC vendors' business :)
00:44:42dattyTorne: Ahhhh
00:44:45dattywikipedia is your friend
00:45:13dattyso the SOC on the Fuze is proprietary SanDisk stuff?
00:45:25TorneNo, it's a SoC mad by Austria Micro Systems
00:45:32TorneHence AMS Sansa
00:45:41TorneThe DAP manufacturers don't generally design their own SoCs
00:45:51Tornebut there are a dozen or more ARM SoC vendors
00:46:02TheSevenwell, besides apple which seems to never have used a stock soc :-)
00:46:07TorneARM only sell the intellectual property
00:46:20Torne(and their build/debug/simulation tools)
00:46:25 Quit LambdaCalculus37 (Read error: 145 (Connection timed out))
00:46:51dattyso i assume it's been looked into all methods of interfacing with the SOC besides the normal USB use?
00:47:12TornePeople have generally looked for anything that's plausibly a JTAG port on our targets, yes
00:47:16Tornethere's not really any method other than JTAG :)
00:47:29Tornethere are a few with a known or suspected JTAG port.
00:47:30dattyim curious how old are you people
00:48:10Torne26, but if you wanna hang out and chat go to #rockbox-community :)
00:48:12TheSevenwell, how old are you? I'm 19 years...
00:48:28dattyi'm still a senior in undergrad so i have some exposure to this stuff but very limited
00:48:49dattyoh ha not for a social reason just to get an idea of your knowledge-to-age ratio of sorts
00:49:11dattyfor example JTAG to me reminds me of the cable we used to program a Xilinx FPGA in one of my classes
00:49:30Tornedatty: #rockbox is logged for people to review/search later, so we try to keep it to on-topic technical discussion
00:49:32TheSeven...which isn't really rockbox-related, hence the link to #rockbox-community
00:50:01dattyo ok sorry bout that
00:50:04 Quit gevaerts (Read error: 60 (Operation timed out))
00:51:09TheSevenkugel: still around?
00:51:11***Saving seen data "./dancer.seen"
00:51:20TheSevenor anybody else who knows how UI drawing works?
00:51:26 Join gevaerts [0] (n=fg@rockbox/developer/gevaerts)
00:51:48*JdGordon| is
00:52:17dattyso is a JTAG "port" for the Fuze the equivalent to the Xilinx EDK/ISE I used to interface using JTAG to the FPGA?
00:52:33dattynm i'm in the other room
00:52:37 Part datty
00:52:39TheSevendo you have any explanation as to why scrolling in the top-level database menu where you can select "artist", "album" etc. could be far slower than e.g. the main menu?
00:53:12 Part toffe82
00:54:05JdGordon|you mean verticle scrolling right?
00:54:36JdGordon|hang on... is the top level one any slower than going into any of the sub items?
00:54:50kugelTheSeven: all menu use the same code basis
00:55:04JdGordon|the database is not the menu :)
00:55:21JdGordon|it uses tagtree which i have no idea how it popuilates the list
00:55:25TheSevenno, it seems to be everything that contains runtime-generated strings, but for some reason also the database top-level menu
00:55:48JdGordon|runtime generated meaning translated? or?
00:56:14TheSeventhings like the main or debug menu are fast, things like the file browser are slow
00:56:19TheSeven(but not accessing the disk)
00:56:58 Join uflops [0] (
00:57:32JdGordon|what about inside settings? like the volume selection list?
00:57:39TheSeventhese are all fine
00:57:51*TheSeven suspects it's scrollable vs. nonscrollable in face
00:58:16TheSeven(scrollable in terms of horizontal here)
00:58:37JdGordon|is there something wrong with your thread switching code maybe?
00:58:54TheSeveni don't think so, that's generic arm code, but of course it causes icache misses
00:59:54JdGordon|very curious indeed :p
01:00:02TheSevenhmm, no, there are horizontally scrolling menus, too, which are also fast
01:01:15JdGordon|the problem is slow resposne going up and down though? no problems if you arnt moving the selection?
01:01:53TheSevenwell, if i aren't moving the selection there's nothing to judge on
01:02:26TheSevenbut if you mean that i stay inside the visible range so that it doesn't need to scroll, that doesn't help
01:02:44TheSevenit even happens with menus that don't have enough entries to scroll at all, but in theory could have more
01:03:32 Quit StealthyXIIGer (Read error: 110 (Connection timed out))
01:03:46 Join StealthyXIIGer [0] (
01:04:22JdGordon|yeah, I have no idea... I was going to suggest that the button driver is sending too many scroll events, but that doesnt make sense either
01:05:34 Quit GeekShadow (Read error: 104 (Connection reset by peer))
01:10:44 Join adiroiban [0] (
01:11:22 Quit mikroflops (Read error: 110 (Connection timed out))
01:13:35 Join faemir [0] (n=faemir@
01:17:19 Quit jdii (Read error: 110 (Connection timed out))
01:18:25 Join sbhsu [0] (
01:18:45kugelTheSeven: the top level database menu is also generated at runtime, it can be customized using a tagnavi.config file
01:19:17kugelwell, everything is generated at runtime, due to localization, but this seems to be different it seems
01:19:38 Join DerPapst [0] (
01:19:59TheSevenlocalization seems to be done through some language database behind the id-based strings
01:20:06TheSeventhe pointer-based strings seem to be the issue
01:21:18kugelyou could try if the "Load to RAM" option in the database settings makes things better or worse
01:21:32TheSevendoesn't seem to change anything
01:29:26 Quit robin0800 (Read error: 110 (Connection timed out))
01:31:17TheSevenbertrik: I have absolutely click-free sound now without pcmbuf changes :-)
01:43:29kugelhow? :)
01:43:50 Join Sajber^ [0] (
01:44:12TheSevenkugel: by splitting the pcmbuf's packets into smaller ones again
01:44:29TheSevenand an extremely streamlined FIQ handler
01:44:41 Quit Sajber^ (Read error: 104 (Connection reset by peer))
01:44:47kugelsounds good
01:45:53 Join Sajber^ [0] (
01:47:40CIA-85New commit by theseven (r23324): Finally a click, pop and other distortion free PCM driver for that fishy bufferless S5L870x I2S controller.
01:48:18TheSeventhis one took far too long. I should have went into that direction from the start
01:50:34CIA-85New commit by theseven (r23325): Implement iPod Nano 2G storage active indicator
01:50:56bertrikTheSeven, looks quite hacky ....
01:51:11 Quit bertrik ("zZz")
01:52:02TheSevenyes, it *IS* hacky, but it at least works finally
01:52:20TheSevenI guess I'll need to spend a few more comments there...
01:53:52TheSevenWTF, why does the bootloader link PCM!?
01:54:36TheSevenoh, funny way how these references sneaked in
01:58:09 Quit Sajber^ ("Leaving.")
01:58:55CIA-85New commit by theseven (r23326): Fix red
02:00:28TheSevendoes anybody have an idea how to get rid of that yellow?
02:00:58TheSevenit's just a "warning: ‘dma_callback’ defined but not used", which is in fact used from ASM
02:05:15CIA-85New commit by theseven (r23327): Hide yellow
02:06:18 Join Sajber^ [0] (
02:08:10kugelTheSeven: (void)dma_callback?
02:08:22TheSevenit's a function
02:08:35TheSevenbut __attribute__((unused)) did the trick
02:09:01kugelyou can also un-static it
02:09:24TheSevenwell, it's local to that file, and has an ambiguous name in other contexts
02:09:35kugelinteresting that gcc doesn't see it
02:09:47TheSevengcc can't see it, as it's inline ASM
02:09:51kugelTheSeven: ______dma__callback then :D
02:10:31TheSevenok, nice and green again
02:11:18kugel__attribute__((unused)) is probably the best way. in fact, I would like to see that used more instead of (void)foo all over the place
02:11:50TheSeveneven though it's a little misleading, as it is in fact used :-)
02:12:29TheSevenok, that was the last issue stopping us from boosting!
02:14:13kugelyou could've also passed dma_callback as "parameter" the inline asm right?
02:14:30TheSevenmy nano is just playing fine with an average ADC reading of 16mA, which is probably too high, as both the ADC and CPU are consuming more power than usual while the measurements are being done
02:14:43 Quit killan (Read error: 104 (Connection reset by peer))
02:14:48TheSevenkugel: yes, I could have voided it in there :-)
02:16:17TheSevenso we probably already exceed apple's battery life!
02:16:55TheSevenI'll drop a patch for that into flyspray and then go to bed
02:17:43kugelI mean like "ldrne r0, %[p]"...: [p]"r"(dma_callback)", but anyway
02:18:06kugelTheSeven: what do you mean by boosting, weren't you running at 192 already all the time?
02:19:50 Join killan [0] (
02:20:55TheSeveni mean going down to 47923200 and boosting only when neccessary
02:21:03TheSevengets us 8mA!
02:21:17 Join peter__b [0] (n=peter_b@
02:21:19 Join Blue_Dude [0] (n=chatzill@rockbox/developer/Blue-Dude)
02:21:52Blue_DudeTheSeven: you got it to work? Awesome! What did the trick?
02:22:05TheSevensplitting down the buffers even further
02:22:21TheSeveni couldn't use your idea because the PCM api in between wouldn't have let me
02:22:34TheSevenwe always forget that there are also other audio sources than pcmbuf :-)
02:22:54Blue_DudeWhatever works. That's great.
02:23:16TheSevenright now, if i get passed a buffer that's at least 2048 bytes, i'll split off the last 1024 bytes
02:23:28 Join Feanor_ [0] (n=Feanor_@
02:23:36Blue_DudeIt's probably also a lot cleaner than hacking up pcmbuf for a single target.
02:23:59TheSeventhis buffer will be requested while the last 1024 bytes of the buffer before are playing, and it will be started as soon as these are done
02:24:25TheSevenwhen the next DMA int comes in, i still have 1024 bytes which i can throw into the dma, while requesting the next one, and so on
02:24:34Blue_DudeSo you roll your own prebuffer while callback is running. That's nifty.
02:24:42TheSevenand i doubt any codec will ever be able to overwrite these 1024 bytes fast enough...
02:25:24TheSeven(and even if it would, this would just end up in dcache and not hurt the DMA)
02:26:37TheSevenand of course the other part of that trick is an ASM FIQ handler that will setup the next DMA transfer within 4 instructions no matter what's going on
02:27:19Blue_DudeI dug some more into the low level pcm drivers this afternoon and I think it might have been possible to hack pcmbuf, but it wouldn't have been pretty. Doing it within your own black box is better though.
02:27:47TheSeventhe only case that can hurt is if there are chunks < 2048 bytes, as these will be directly used as the "tail buffers", so the next one will be requested immediately, which may lead to codecs overwriting them ahead of time again, but this could possibly still be caught by dcache
02:28:11 Quit DerPapst ("Leaving.")
02:28:14TheSeven(which is 4KB on that target)
02:28:37Blue_DudeWhat happens when the codec stops feeding data at the end of the playlist? You will get the tail end then. However, the codec won't be overwriting either.
02:29:55 Quit peter-b (Read error: 145 (Connection timed out))
02:30:02TheSeventhis will just mean that the pcmbuf can't supply data while I'm playing the tail, so it will not schedule a new buffer for pushing to the dma, and as soon as the DMA of the tail finishes, it will realize there is nothing more to do and call that stop callback
02:31:07TheSevenlinuxstb: we seem to have reached apple's battery runtime with some tricks :-)
02:31:07Blue_DudeOK, just so long as your prebuffer doesn't choke on a smaller than expected payload. That's just details though. Congratulations.
02:31:27linuxstbTheSeven: Nice. Is battery_bench working now?
02:31:44TheSevenlinuxstb: it's failing every night because of a new reason
02:32:06TheSevenBlue_Dude: it shouldn't choke on that, as it's able to chain tails directly if they're too small to split them
02:33:00Blue_DudeTheSeven: that's excellent.
02:33:06TheSeventhis ranges from FTL problems to stack overflows to whatever
02:33:24TheSevenI won't do it tonight because my battery isn't really full right now
02:33:42TheSevenbut i'll post the patch to flyspray right away
02:43:22TheSevenFS #10714
02:44:33TheSevenBlue_Dude: BTW I spotted two things that look like bugs in pcmbuf:
02:45:37TheSevenline 1029: The actual buffer usage is higher than that, as some date (which isn't included in that value) may still be playing
02:45:48Blue_DudeTheSeven: that doesn't surprise me at all.
02:46:08TheSevenand line 1041: I don't see what this is supposed to do. Looks like someone copy-pasted something in the wrong place.
02:46:31TheSevenwhy for heaven's sake does it divide a percentage value by pcmbuf_unplayed_bytes?
02:47:00TheSevenshouldn't that be pcmbuf_size?
02:47:01Blue_Dudere line 1029: that's pretty approximate because it doesn't include pcm_unplayed_bytes or whatever it is from pcm.c
02:48:49Blue_Dudere line 1041: you're probably right. I'm not 100% sure where those functions are called from, but they smell like debug screen info to me
02:49:14 Quit Rondom (Nick collision from services.)
02:49:25 Join Rondom [0] (
02:51:12***Saving seen data "./dancer.seen"
02:53:42Blue_Dudere line 1029: the problem is that pcmbuf_unplayed_bytes is decremented as soon as the chunk is sent to the DMA, not when it's played. If you want to be really accurate, you'd have to keep add last_chunksize to pcmbuf_unplayed_bytes. That assumes that the entire chunk is off limits whether part of it has been played or not, which is safest.
02:54:22TheSevendepends on what this function is good for
02:54:52TheSeventhis could either be space used in terms of allocated/locked, or in terms of data being in the buffer ready to be retrieved
02:54:59Blue_DudeAs I said, it smells like howgozit info vs. operational necessity.
02:55:37Blue_DudeIf you're looking for free buffer space, you have to assume that the currently playing chunk is not free though.
02:55:37TheSevenbtw, that pcmbuf_mix_free is only used by pcmbuf itself
02:56:26TheSeventhe same is true for the other one
02:56:43TheSevenand it's both on line 882
02:56:57TheSeven"A get-it-to-work-for-now hack" :-)
02:57:05Blue_DudeI have been poking around in pcmbuf and playback for a few weeks with an eye towards tightening them up.
02:57:15Blue_DudeYeah, that's right. :)
02:57:34Blue_DudeSo some of this stuff may not last the winter.
02:58:09 Quit liar (Read error: 148 (No route to host))
02:58:45*TheSeven will go to bed now and continue to poke on this "sluggish UI" issue tomorrow
02:59:14Blue_DudeIt's not just you. The UI really started to drag a few weeks back.
02:59:45Blue_DudeI've got to go pack. Back next week. Enjoy!
03:00:50Blue_DudeOh yeah. Line 882: voice buffer stuff. That will be broken out.
03:02:49 Quit Sajber^ (Read error: 104 (Connection reset by peer))
03:03:37 Quit uflops (Remote closed the connection)
03:03:45 Quit Blue_Dude ("ChatZilla 0.9.85 [Firefox 3.5.3/20090824101458]")
03:04:30 Quit froggyman ()
03:05:36 Quit amiconn (Nick collision from services.)
03:05:39 Join amiconn_ [0] (i=quassel@rockbox/developer/amiconn)
03:05:58 Nick amiconn_ is now known as amiconn (i=quassel@rockbox/developer/amiconn)
03:07:31 Join mikroflops [0] (
03:08:05 Join pixelma_ [0] (i=quassel@rockbox/staff/pixelma)
03:08:05 Quit pixelma (Nick collision from services.)
03:08:24 Nick pixelma_ is now known as pixelma (i=quassel@rockbox/staff/pixelma)
03:09:58 Quit mikroflops (Remote closed the connection)
03:10:33 Join mikroflops [0] (
03:14:08 Quit Feanor_ (Remote closed the connection)
03:17:39 Quit mikroflops (Remote closed the connection)
03:22:18 Join mikroflops [0] (
03:23:15 Join Guest61935 [0] (
03:23:15 Quit Guest61935 (Client Quit)
03:28:48 Join uflops [0] (
03:30:18 Quit mikroflops (Read error: 54 (Connection reset by peer))
03:32:50 Join petur [50] (n=petur@rockbox/developer/petur)
03:44:40 Quit petur ("later")
04:02:56 Join Hillshum [0] (
04:13:22 Quit TheSeven (Nick collision from services.)
04:13:43 Join The_Seven [0] (n=theseven@rockbox/developer/TheSeven)
04:13:55 Nick The_Seven is now known as TheSeven (n=theseven@rockbox/developer/TheSeven)
04:18:46 Quit kugel (Remote closed the connection)
04:23:08 Quit adiroiban (Read error: 110 (Connection timed out))
04:28:27 Join BHSPitLappy [0] (n=BHSPitLa@unaffiliated/bhspitmonkey)
04:33:00Unhelpfulkugel: i was using target="arm-elf-eabi" in i also had added support for building binutils snapshots. here's a completely untested update. you will need to provide the gcc patch yourself. :)
04:33:06Unhelpful <- grr.
04:43:45 Join froggyman [0] (
04:51:13***Saving seen data "./dancer.seen"
04:54:40 Quit Strife89 (Read error: 104 (Connection reset by peer))
04:55:02 Join Strife89 [0] (
05:01:10 Quit panni_ (Read error: 104 (Connection reset by peer))
05:04:19 Quit froggyman ("CGI:IRC")
05:29:13 Quit MethoS- (Remote closed the connection)
05:29:16 Join jdii [0] (
05:37:25 Join Strife1989 [0] (
05:52:51 Quit Strife89 (Read error: 113 (No route to host))
05:57:25 Quit TheSeven (Nick collision from services.)
05:57:46 Join The_Seven [0] (n=theseven@rockbox/developer/TheSeven)
05:57:58 Nick The_Seven is now known as TheSeven (n=theseven@rockbox/developer/TheSeven)
06:07:12 Join lifeless_ [0] (n=lifeless@
06:17:14 Quit SUSaiyan (Read error: 54 (Connection reset by peer))
06:21:26 Quit TheSeven (Nick collision from services.)
06:21:47 Join The_Seven [0] (n=theseven@rockbox/developer/TheSeven)
06:21:59 Nick The_Seven is now known as TheSeven (n=theseven@rockbox/developer/TheSeven)
06:22:57 Join SUSaiyan [0] (
06:24:36 Join tomers [0] (
06:25:27 Quit lifeless__ (Read error: 113 (No route to host))
06:30:50 Quit tomers ("ChatZilla 0.9.85 [Firefox 3.0.14/2009090216]")
06:51:13 Quit Strife1989 ("My number of files is OVER 9000!")
06:51:14***Saving seen data "./dancer.seen"
07:04:29 Join Riku [0] (
07:07:06 Quit Riku (Read error: 104 (Connection reset by peer))
07:16:50CIA-85New commit by tomers (r23328): Minor Hebrew translation fixes
07:18:00 Join tomers [0] (
07:18:10tomersDoes anyone knows if it possible to cross-compile rbutil to Win32 from Linux?
07:23:39 Quit Lss (Read error: 110 (Connection timed out))
07:29:52 Quit jdii (Read error: 110 (Connection timed out))
07:33:57 Join Horschti [0] (n=Horscht2@xbmc/user/horscht)
07:40:48 Join FOAD_ [0] (
07:46:10 Join adiroiban [0] (
07:52:29 Quit Horscht (Read error: 110 (Connection timed out))
07:56:15 Quit adiroiban ("Leaving.")
07:56:32 Join adiroiban [0] (
07:56:45 Part scp1
07:56:57 Quit FOAD (Read error: 110 (Connection timed out))
07:56:57 Nick FOAD_ is now known as FOAD (
07:58:54 Join AJ707 [0] (
08:10:33 Part AJ707
08:19:53 Quit adiroiban (Read error: 110 (Connection timed out))
08:20:56 Quit StealthyXIIGer (Read error: 110 (Connection timed out))
08:27:20 Join robin0800 [0] (n=quassel@
08:30:54 Join n1s [0] (n=n1s@rockbox/developer/n1s)
08:39:55 Join Horscht86 [0] (
08:47:56 Join Rob2222 [0] (
08:51:17***Saving seen data "./dancer.seen"
09:04:19 Quit Horschti (Read error: 110 (Connection timed out))
09:04:46 Quit Hillshum ("Ex-Chat")
09:06:06 Quit Rob2223 (Read error: 110 (Connection timed out))
09:06:17 Quit BHSPitLappy ("Ex-Chat")
09:07:09 Join FlynDice [0] (
09:30:29 Quit JdGordon| (Read error: 110 (Connection timed out))
09:32:54 Join stoffel [0] (
09:37:52 Quit tomers ("ChatZilla 0.9.85 [Firefox 3.0.14/2009090216]")
09:42:35 Nick uflops is now known as mikroflops (
09:47:56 Join Rob2223 [0] (
09:49:58 Join esperegu [0] (n=quassel@
09:51:45 Quit n1s (Read error: 110 (Connection timed out))
09:56:26 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
09:56:27 Join n1s [0] (n=n1s@rockbox/developer/n1s)
10:03:06 Quit Rob2222 (Read error: 110 (Connection timed out))
10:09:36 Join BHSPitLappy [0] (n=BHSPitLa@unaffiliated/bhspitmonkey)
10:21:32 Quit evilnick (Read error: 113 (No route to host))
10:22:10 Nick Horscht86 is now known as Horscht (
10:22:32 Quit phanboy4 ("Leaving")
10:45:08 Quit faemir ("Leaving")
10:48:34 Join GeekShadow [0] (n=Antoine@reactos/tester/GeekShadow)
10:51:21***Saving seen data "./dancer.seen"
11:13:12 Join GeekShado_ [0] (
11:14:25 Join ender` [0] (
11:21:06 Join gtkspert [0] (
11:24:20 Join pamaury [0] (
11:29:27 Quit GeekShadow (Read error: 113 (No route to host))
11:36:23 Quit gtkspert_ (Read error: 101 (Network is unreachable))
11:36:58*TheSeven is confused by a build that worked flawlessly yesterday, and today it doesn't even boot any more
11:48:13 Quit linuxstb (Remote closed the connection)
11:52:50 Join flydutch [0] (
11:53:22 Join Paprica [0] (
11:55:43Papricahi, i first time compile in linux (i use the guide in the docs) and i have a problem
11:55:56Papricamy "making" look like that
11:56:10Papricawhat i need to do?
11:57:21n1sconfigure should turn that warning off, it's basically harmless; are you using a recomended compiler version?
11:59:12Papricagood question i dont know the answer XD i just follow the instructions here
11:59:26n1sor did you run configure without the compiler in the path maybe? if so rerun configure
12:00:21Papricai did conigure
12:01:41n1sconfigure should tell you if it can't find the compiler
12:02:05TheSevenn1s: this looks like there is another arm-elf-gcc it found that isn't ours
12:02:25PapricaTheSeven what should i do?
12:02:39TheSevenput the right one to the path
12:02:57TheSevenwhat does a "which arm-elf-gcc" result in?
12:04:34TheSevenlooks sane
12:04:59Papricaso? :\
12:05:05TheSevenarm-elf-gcc −−version
12:05:35 Join tomers [0] (
12:06:05n1sTheSeven: yeah
12:06:08n1sit can also happen if you un confgure without the right crosscompiler in the path as it then can't determine the version and make the choices about disabling warnings etc.
12:06:23TheSevenand another thing you can check: ls -l /usr/local/arm-elf/bin/arm-elf-gcc
12:06:33TheSevendoes that show the date you ran
12:07:15Papricaarm-elf-gcc (GCC) 4.0.3
12:07:15PapricaCopyright (C) 2006 Free Software Foundation, Inc.
12:07:15PapricaThis is free software; see the source for copying conditions. There is NO
12:07:15DBUGEnqueued KICK Paprica
12:07:15Papricawarranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
12:07:35TheSevenlooks sane, too
12:08:17TheSevennow check that date, and if that looks sane, too, run "make reconf && make" and see if those warnings still show up
12:08:37tomerszagor, badger: ping
12:09:09PapricaTheSeven, -rwxr-xr-x 2 root root 154957 2009-10-24 11:20 /usr/local/arm-elf/bin/arm-elf-gcc
12:09:44TheSevenok, so the right compiler is in the path
12:10:28Papricaseem to be ok :)
12:10:31Papricathank you :P
12:11:57TheSevenok, so the first quess was right: [11:58]<n1s>or did you run configure without the compiler in the path maybe? if so rerun configure
12:12:58n1sTheSeven: the fun thing is that configure prints a big warning when you run it without the compiler in the PATH...
12:13:44n1salso, slightly related; i wonder if it would make sense to fix those pointer signedness mismatches
12:14:38TheSevenyes, I wondered about that, too
12:15:01TheSevenwhile most of them probably can't, some of them may be potentially harmful
12:15:40TheSevenand I don't really like a "disconnect the light" strategy either
12:16:27n1snah, it's just that they all showed up with the last gcc upgrade
12:21:25CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
12:21:25*TheSeven wonders why pointers should ever be signed
12:21:26Papricaanother question about linux, which editor do you use?
12:22:06 Join Utchybann [0] (
12:24:29 Quit pamaury (Read error: 113 (No route to host))
12:27:49*TheSeven wonders what an #ifdef CPU_PP inside an #if defined(CPU_PP) is good for
12:28:16UtchybannTheSeven: mpegplayer seems to play video (and sound) way too fast with r23328.
12:28:30TheSevenwhen did this start to happen?
12:30:03TheSevencan somebody else have a look at config.h:775 and see if I'm just blind or this is really just confusing nonsense that's being done there?
12:30:32UtchybannI just compile r23328 and try to play elephantdream video (from rockbox site).
12:31:18TheSevenhow does the sound behave? squirrel mode? or just skipping?
12:32:41Utchybannsquirrel mode? It is like play the video at 100fps.
12:33:54Utchybanna timer/clock change?
12:37:04 Join DerPapst [0] (
12:37:18 Join kugel [0] (n=kugel@rockbox/developer/kugel)
12:37:20n1sTheSeven: yeah that #ifdef mess in config.h looks rather odd
12:38:17 Quit lifeless_ (Remote closed the connection)
12:41:42UtchybannTheSeven: mp3/ogg are playing fine without click/pop. The problem seems limited to mpegplayer. metronome runs fine.
12:43:00 Quit kugel (Remote closed the connection)
12:44:30n1shmm, this also means that all pp targets use the SW_CORELOCK workaround for the broken swap instruction... I wonder if that's correct
12:45:24amiconnit is
12:45:35amiconn(iirc - search the log for details)
12:46:03amiconnjhMikeS worked on this
12:46:32Bob_CThe endif comments certainly don't match the #if blocks
12:47:12n1samiconn: just the #ifdef logic that's a bit weird then
12:47:38 Join bertrik [0] (
12:48:26Bob_CWould it be considered bad style to indent #if blocks to make the nesting clearer?
12:48:59amiconnChecking the history of config.h will probably help figuring things out properly
12:49:40amiconnBob_C: Imho preprocessor statements should never be indented
12:50:05Bob_CWhy not?
12:50:28amiconnI looks all wrong to me... no technical reason
12:50:37*TheSeven already found a place where they *are* already indented, but I can't remember where that was
12:50:39 Join kugel [0] (n=kugel@rockbox/developer/kugel)
12:50:52bertrikI like to see the original condition repeated in the #endif if the #ifdef / #endif is very long or highly nested
12:51:01TheSevenin files that just consist of preprocessor statements, it may in fact make sense to indent them
12:51:26***Saving seen data "./dancer.seen"
12:51:33amiconnbertrik: That's usually the case, just in the mentioned case it's all broken
12:51:37bertrik(I mean an inline comment added to the #endif)
12:51:44TheSevenbertrik: #else /* ... */ is also a mess
12:52:02TheSevensometimes the original conditions are in there, sometimes the condition on which the else block is used
12:59:43TheSevenline 808 also looks very weird to me
13:00:11 Join funman [0] (n=fun@rockbox/developer/funman)
13:00:28TheSevenif this should ever be reached, it will clash with line 786
13:01:28*TheSeven also just realized that mutexes aren't interrupt-safe
13:03:06bertrikno, I think you can't use a mutex at all in interrupt context
13:03:36TheSevenbut they are being used in an interrupt context
13:03:43bertrikyou can post a message in a queue, but that's basically the only kernel primitive you can call, I think
13:03:58TheSeven...and hopefully wakeups
13:04:08bertrikyeah, those too :)
13:04:43funmanmutexes are used in interrupt context ?
13:04:45amiconnYou can't block on a mutex in interrupt context, that would cause a freeze if it actually blocks
13:05:14TheSevenof course, and I'm currently trying to track down a place where this happens
13:05:35bertrikWe should have some way to panic on that I think
13:05:37amiconnThis is obvious - the thread holding the mutex can't let it go because it won't execute unless the isr returns, which it can't do because of the block
13:05:44amiconnThis is by design
13:06:09funmanbertrik: how we know if we're in an isr is platform dependant
13:06:34amiconnWhy would we need a panic for this?
13:06:35bertrikgreat, so we can implement it for some platforms :)
13:07:03kugelthe comment makes me wonder why NOCACHEBSS_ATTR is only defined for PP
13:07:08funmanwell we can check the cpsr for interrupt state
13:07:29amiconnfunman: Coldfire, SH1, MIPS, ...
13:07:43funmanTheSeven: what about printing the current stack if such a thing happen and look for addresses ?
13:07:55funmanamiconn: yeah it's platform dependant, but this could help TheSeven on nano2g
13:08:16TheSevenI think I already know what's going on, but need to double-check that
13:08:18 Join evilnick [0] (
13:08:26bertrikamiconn, it's an obvious bug if this happens, so I think it's better to show this as clearly as possible
13:08:28 Join kugel_ [0] (
13:08:40 Quit kugel (Nick collision from services.)
13:08:44 Nick kugel_ is now known as kugel (
13:09:12 Quit tomers (Read error: 60 (Operation timed out))
13:10:10bertrikmaybe only panic when the kernel is compiled in some kind of debugging mode to avoid slowdowns for code that already does things correctly
13:10:39TheSevenbertrik: there already is a flag for that, but it's barely used
13:11:04funmanDEBUG ?
13:11:11amiconnI guess it depends on how expensive that check is
13:11:32TheSevenfunman: no, there's something specific for the kernel
13:12:11bertrikDEBUG enables KERNEL_OBJECT_CHECKS
13:15:02bertrikamiconn, what do you think? Would you approve of extra checking in the kernel (#ifdef DEBUG or similar) that verify if a call is made in the right context?
13:17:54 Join linuxstb [0] (n=linuxstb@rockbox/developer/linuxstb)
13:18:34 Quit togetic (Read error: 110 (Connection timed out))
13:24:34 Quit Utchybann (Read error: 113 (No route to host))
13:24:34bertrikfunman, I think you had a patch for more verbose panic for the ams sansas, right? Or is it already in?
13:25:42funmani have posted a patch to get the address causing the fault from cp15 register
13:26:02funmanit's not in because we need to check every CPU which implement this (FAR = Fault Address Register)
13:26:23funmanit only gives you more details on Data Aborts
13:28:45bertrikso basically you just need a list of possible targets where we could use this and have people test it?
13:29:03funmanno, we would know from the documentation
13:29:13 Quit robin0800 (" - Chat comfortably. Anywhere.")
13:29:36 Join robin0800 [0] (
13:29:49 Quit robin0800 (Client Quit)
13:30:13funmanTorne explained it on FS #10296 : it's available on ARM CPUs with MMU or MPU. The ARM7TDMI (PP) do not
13:30:22 Join robin0800 [0] (n=quassel@
13:32:39kugelamiconn: have you seen my experiments with using the F/X lcd_bitmap_transparent_part() asm optimization for other arm targets?
13:32:46bertrikfunman, so what's stopping you from implementing it?
13:33:36funmanbertrik: lack of usefulness, I have used this patch for my own needs but right now i don't need an extended data abort handler
13:34:06 Quit robin0800 (Client Quit)
13:34:30funmanI think it would require adding yet another feature like #define HAVE_ARM_AND_MMU_OR_MPU
13:35:45TheSevensomeone around who wants to play a little with nano2g undervolting?
13:35:55 Quit BHSPitLappy (Remote closed the connection)
13:36:50 Join robin0800 [0] (
13:38:48kugelTheSeven: does the datasheet give any recommendations?
13:39:09TheSevenwhich datasheet, and recommendations for what?
13:39:28kugelabout voltages
13:39:45TheSevenif what the DS says is right, apple is already undervolting it
13:39:58 Quit yosafbridge ("Coyote finally caught me")
13:40:00kugelyea, the sandisk Of also does it
13:40:07 Join yosafbridge [0] (
13:40:25TheSeventhey're doing 1075mV while samsung recommends 1100-1300
13:40:43TheSeveni'm doing 1000 at 200MHz and 900 at 50MHz and it's running fine
13:41:03kugelI probably wouldn't go below applies setting otherwise it may fail on some devices
13:41:16bertrikIt could be of course that 1 in 10 players can not go that low reliably
13:41:32TheSevenyes, and that's why i want people to test this
13:41:56 Join tomers [0] (n=chatzill@
13:42:04TheSevenin fact i could even go to 975 (200MHz) and 800 (50MHz) from the CPU side
13:42:11TheSevenbut the quartz osc doesn't seem to like the 800
13:45:33 Quit BHSPitMonkey (Read error: 113 (No route to host))
13:46:31 Quit JackWinter (Remote closed the connection)
13:47:40bertrikTheSeven, having more data and knowing where the limits are, is always good of course :)
13:48:06 Join JackWinter [0] (
13:48:08TheSeventhat's why there is FS #10714
13:58:19 Join MethoS- [0] (n=clemens@
14:05:03 Quit antil33t (Read error: 54 (Connection reset by peer))
14:05:09 Join antil33t [0] (n=Mudkips@
14:07:12 Quit flydutch ("/* empty */")
14:16:13 Quit MethoS- (Remote closed the connection)
14:17:16 Nick YPSY is now known as Ypsy (n=ypsy@
14:22:05 Join efyx_ [0] (
14:26:45 Quit funman ("free(random());")
14:26:50 Join Crackerizer [0] (
14:27:44CrackerizerI just finished diacritic patch and need reviews.
14:43:12 Join kyle6513 [0] (n=kyle@
14:44:19 Quit kyle6513 (Client Quit)
14:51:28***Saving seen data "./dancer.seen"
14:57:54 Join gtkspert_ [0] (
15:00:42 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
15:02:16 Join dfkt_ [0] (i=dfkt@unaffiliated/dfkt)
15:05:10 Nick chaos is now known as chaos_ (n=ch4os@gentoo/user/ch4os)
15:05:37 Quit dfkt (Nick collision from services.)
15:05:40 Nick dfkt_ is now known as dfkt (i=dfkt@unaffiliated/dfkt)
15:05:42 Nick chaos_ is now known as chaos (n=ch4os@gentoo/user/ch4os)
15:10:07bertrikI made some very limited progress on the meizu m6 sp display last night
15:10:33bertriknow I can also get random vertical stripes on the screen and not just horizontal
15:10:58bertrikit looks like the screen is refreshed from left to right instead of top to bottom
15:12:01bertrikAlso I realised that reading an LCD is basically impossible because the SPI interface towards the LCD lacks a MISO pin (at least if the schematic I have is representative for my player)
15:12:40bertrikThe OF seems to decide at runtime between two drivers, based on the LCD id read back from the LCD
15:13:41 Quit gtkspert (Read error: 101 (Network is unreachable))
15:13:56 Join bertrik_ [0] (
15:14:00 Quit bertrik ("Leaving")
15:14:11 Nick bertrik_ is now known as bertrik (
15:17:18 Quit meermanr (Remote closed the connection)
15:19:28 Join dfkt_ [0] (i=dfkt@unaffiliated/dfkt)
15:19:45 Quit dfkt (Nick collision from services.)
15:19:48 Nick dfkt_ is now known as dfkt (i=dfkt@unaffiliated/dfkt)
15:26:42amiconnbertrik: Many lcd controllers allow to change the update direction
15:28:11amiconnIf the meizu m6sp lcd doesn't allow this, or for some reason it's unedesirable, it will profit from kkurbjun's vertical 16 bit driver
15:33:25kugelamiconn: have you seen my experiments with using the F/X lcd_bitmap_transparent_part() asm optimization for other arm targets?
15:37:50 Quit tomers ("ChatZilla 0.9.85 [Firefox 3.5.3/20090824101458]")
15:38:05 Quit Horscht ("Verlassend")
15:40:07 Quit JackWinter (Read error: 104 (Connection reset by peer))
15:41:04 Join JackWinter [0] (
15:42:31 Join dfkt_ [0] (i=dfkt@unaffiliated/dfkt)
15:46:11 Quit JackWinter (Remote closed the connection)
15:47:05 Join JackWinter [0] (
15:47:30 Quit T44 (Read error: 110 (Connection timed out))
15:53:20 Quit dfkt (Nick collision from services.)
15:53:23 Nick dfkt_ is now known as dfkt (i=dfkt@unaffiliated/dfkt)
15:55:45 Quit JackWinter (Remote closed the connection)
15:56:28 Join JackWinter [0] (
15:59:07 Join Horscht [0] (n=Horscht2@xbmc/user/horscht)
16:00:02 Join pamaury [0] (
16:03:41 Join tomers [0] (
16:05:10 Quit Horscht ("Verlassend")
16:09:44 Quit JackWinter (Remote closed the connection)
16:15:20 Join JackWinter [0] (
16:21:21 Quit pamaury ("exit(*(int *)0 / 0);")
16:25:02 Join Horscht [0] (n=Horscht2@xbmc/user/horscht)
16:26:38linuxstbkugel: Where are those experiments?
16:30:54 Quit Paprica (Read error: 110 (Connection timed out))
16:31:03 Join Paprica [0] (
16:32:16kugelbasically, the F/X optimization is quite generic and can be used on all color display arm targets
16:37:36 Join Strife89 [0] (
16:39:13 Quit TheSeven (Read error: 145 (Connection timed out))
16:42:18 Quit esperegu (Read error: 104 (Connection reset by peer))
16:43:48 Quit tomers (Read error: 113 (No route to host))
16:46:54 Join dfkt_ [0] (i=dfkt@unaffiliated/dfkt)
16:48:43 Quit dfkt (Nick collision from services.)
16:48:46 Nick dfkt_ is now known as dfkt (i=dfkt@unaffiliated/dfkt)
16:51:14 Join dfkt_ [0] (n=dfkt@unaffiliated/dfkt)
16:51:31***Saving seen data "./dancer.seen"
16:52:35 Quit dfkt (Nick collision from services.)
16:52:38 Nick dfkt_ is now known as dfkt (n=dfkt@unaffiliated/dfkt)
16:59:36 Join esperegu [0] (n=quassel@
17:09:10Tornebertrik: I was gonna go ahead and write a better abort handler at some point, and make it conditional on the right things.. there's a few things beyond what funman's patch does you can print, and it doesn't handle armv6 fully
17:09:19Tornebertrik: if you think it's useful i can bump it up the list ;)
17:13:13 Join StealthyXIIGer [0] (
17:16:42 Quit krazykit ("2.6.31")
17:18:58 Join krazykit [0] (n=kkit@
17:27:40 Quit dfkt (Read error: 110 (Connection timed out))
17:45:12 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
17:47:16 Quit DerPapst ("Leaving.")
17:50:26 Join Ubuntuxer [0] (
17:54:38 Join tvelocity [0] (
17:55:37 Quit StealthyXIIGer (Read error: 104 (Connection reset by peer))
17:55:57 Join StealthyXIIGer [0] (
18:01:35 Join tomers [0] (
18:01:55CIA-85New commit by kugel (r23329): Speed up lcd_update_rect by ~2.5% by counting towards 0 in the loop.
18:03:12 Join toffe82 [0] (
18:05:19tomerszagor, badger: ping
18:08:17 Join panni_ [0] (
18:12:05 Quit Ubuntuxer (Read error: 110 (Connection timed out))
18:13:36 Join adiroiban [0] (n=adiroiba@
18:18:12 Join dfkt_ [0] (n=dfkt@unaffiliated/dfkt)
18:18:20CIA-85New commit by tomers (r23330): scramble: Make function static
18:19:03 Join togetic [0] (n=togetic@unaffiliated/ibuffy)
18:20:29 Quit dfkt (Nick collision from services.)
18:20:34 Nick dfkt_ is now known as dfkt (n=dfkt@unaffiliated/dfkt)
18:23:27 Quit JackWinter (Read error: 104 (Connection reset by peer))
18:25:26 Join JackWinter [0] (
18:31:43 Join NoGare [0] (i=NoGare@
18:42:26CIA-85New commit by FlynDice (r23331): Add #ifdef HAVE_MULTIDRIVE to HS timings switch for uSD cards. ...
18:44:18 Join Topy44 [0] (
18:45:32 Join bertrik_ [0] (
18:45:43 Quit bertrik (Read error: 113 (No route to host))
18:51:32***Saving seen data "./dancer.seen"
18:54:44 Join Sir_Konrad [0] (
18:58:24 Quit NoGare (Client Quit)
18:58:31Sir_KonradiPod shuffle?
19:00:15 Quit robin0800 (Remote closed the connection)
19:03:04krazykitSir_Konrad, what about it?
19:03:11AlexPSir_Konrad: Not supported
19:03:24Sir_Konradthat's all I needed to know. ;P
19:03:38Sir_Konradkthxbye AlexP.
19:03:41 Part Sir_Konrad ("Leaving")
19:08:52 Quit esperegu (Read error: 60 (Operation timed out))
19:09:32 Join esperegu [0] (n=quassel@
19:09:35 Join flydutch [0] (
19:17:48 Join liar [0] (n=liar@
19:20:31 Join funman [0] (n=fun@rockbox/developer/funman)
19:20:48 Quit Crackerizer ("CGI:IRC")
19:21:33 Quit stoffel (Remote closed the connection)
19:23:22 Join TheSeven [0] (n=theseven@rockbox/developer/TheSeven)
19:27:06 Join petur [50] (n=petur@rockbox/developer/petur)
19:32:11 Quit Llorean ("Leaving.")
19:42:45n1skugel: re r23329 i wonder if gcc will not do a similar optimization at higher optimization levels
19:47:13 Join Lss [0] (
19:49:12 Join MethoS- [0] (n=clemens@
19:50:42kugeln1s: we use -O2 already IIRC
19:52:42 Join Vitor1001 [0] (n=vitor@
19:53:29Vitor1001anybody at google mentor summit?
19:53:48kugelpetur and scorche
19:54:08 Join phanboy4 [0] (
19:54:15scorche|shVitor1001: hi
19:54:44gevaertsat least they said so. We have no idea if they're actually there. They might be partying somewhere else for all wek know
19:54:48Vitor1001scorche|sh: Hi
19:55:04Vitor1001scorche|sh: Ronald and me booked a room to discuss multimedia
19:55:18scorche|shVitor1001: today?
19:55:22Vitor1001and not reusing code between ffmpeg and other projects
19:55:42Vitor1001Yes, let me just look at the board which room
19:55:51scorche|shVitor1001: i will check the board shortly
19:56:03scorche|shalso, there is #mentorsummit here on freenode if you wish
19:56:12*gevaerts pretends to misunderstand
19:56:19gevaertsWhat's wrong with reusing code?
19:56:50funmangevaerts: keeping forked branches up to date with each other is difficult and time consuming
19:56:57Vitor1001it's not wrong. It is so right we want to discuss how to do it best ;)
19:57:27Vitor1001el-golea room
19:57:43*petur waves (or tries to if fennec works)
19:57:48scorche|shi will make petur record for the benefit of the more codec-oriented people ;)
19:57:54n1skugel: nope, core is compiled with -O for all arm targets and Os for sh and coldfire
19:59:35gevaertsVitor1001: that's not what you said! You wanted to talk about "not reusing code between ffmpeg and other projects"!
20:00:14Vitor1001gevaerts: Well, was what I understand from what ronald say
20:00:48Vitor1001The spirit is the same
20:01:01Vitor1001It is not a "use our libraries, they are cool" session
20:01:48Vitor1001I think rockbox fixed point wma decoder particularly interesting
20:02:05Vitor1001We'll, we are there already, so if you want to discuss it live, you can just drop in
20:04:30saratogawe can't use ffmpeg directly because ffmpeg doesn't do fixed point
20:05:20funmanwill you talk about possible integration of rockbox modifications into ffmpeg svn ?
20:05:24n1sand we don't do floating point :)
20:11:09 Quit sinthetek (Read error: 110 (Connection timed out))
20:11:09 Nick bertrik_ is now known as bertrik (
20:11:12 Join DerPapst [0] (
20:11:18 Join polobricolo [0] (
20:15:13 Quit jfc^3 (Read error: 54 (Connection reset by peer))
20:16:24 Join jfc^3 [0] (n=john@
20:17:02 Join BHSPitLappy [0] (n=BHSPitLa@unaffiliated/bhspitmonkey)
20:17:28 Quit JackWinter (Read error: 104 (Connection reset by peer))
20:17:28 Join uflops [0] (
20:17:31scorche|shsaratoga: around?
20:17:43 Quit mikroflops (Read error: 60 (Operation timed out))
20:18:24 Quit MethoS- (Remote closed the connection)
20:19:54scorche|shlinuxstb: you as well? ;)
20:22:17*linuxstb is here
20:22:21 Join JackWinter [0] (
20:22:26 Join Buschel [0] (
20:23:30 Join BBB [0] (n=rbultje@
20:23:54saratogascorche|sh: i'm back
20:24:58peturlinuxstb, saratoga: the ffmpeg guys have some questions, hold on
20:25:05 Quit JackWinter (Remote closed the connection)
20:26:25 Join JackWinter [0] (
20:27:00 Join j-b [0] (n=jb@videolan/developer/j-b)
20:27:15BBBsaratoga, how diff is wmafixed from wmafloat/ffmpeg?
20:27:38saratogamost of the basic functions are similar, and it has been back ported to ffmpeg based projects
20:27:55funmansome tables/functions are given section attributes to be put in faster memory
20:28:57saratogawe use a different mdct now, and the decode_superframe function has been split into a function that works on smaller pieces (the original wma codec could operate on enormous blocks of memory at one time IIRC)
20:29:56saratogawe also don't use the ffmpeg parser
20:29:59saratogaasf parser
20:30:06BBBsaratoga, can you submit the superframe decode function split? that would help
20:30:09BBBit's a big ugly function
20:30:22BBBthe fact tha tyou don't use asfdec shouldn't matter much right?
20:30:50saratogalinuxstb wrote it originally IIRC, but I think it shouldn't be too hard to port
20:31:02saratogaalthough i'm not sure if the ffmpeg API would be ok with it
20:33:28saratogai'd have to review the change, but I suppose wma_decode_superframe could be replaced with wma_decode_superframe_frame (e.g. the version that returns exactly 2048 samples) and be made to call wma_decode_superframe_init everytime it finished a superframe
20:35:06BBBcould you submit patches for that?
20:36:37saratogaeventually, but i have little time during the semester
20:36:52saratogawhy the new interest in our wma changes?
20:37:33saratogafunman: unrelated question, what does it mean when the clip crashes by simply powering down, and reboots on any button press at all (not just power)?
20:38:00funmanI'd say panic
20:38:12funmanpanicf() will reboot on any button press
20:39:06saratogashouldn't that print an error?
20:39:36 Quit Vitor1001 (Read error: 110 (Connection timed out))
20:40:13peturit should print stuff and reboot if you hit a key in that screen
20:41:33funmansaratoga: i've seen unexplained reboots already, how can i reproduce this one ?
20:42:24saratogafunman: I get them pretty easily with this patched clip build:
20:42:41saratogatakes maybe 1-2 tracks before it randomly happens while playing
20:42:50saratogaits just the latest buffering patch
20:42:54saratogai'm a little confused whats happening
20:42:58funmando you have lcd always on
20:42:59kugelsaratoga: is the display off so that you can't read the screen?
20:43:20saratogakugel: theres a click sound like the DAC powered off, but its possible the rest of the SOC is still on
20:44:10kugelit sounds like a panicf to me too, you of course need the display on in order to read it
20:45:45funmanif kugel and me agree it surely must be the Truth ;))
20:46:21saratogai'll disable the backlight timer and try again
20:47:33 Quit phanboy4 (Read error: 104 (Connection reset by peer))
20:48:40BBBsaratoga, it'd be nice to integrate it back
20:48:43BBB(sorry for replying slow)
20:49:39 Quit liar (Remote closed the connection)
20:50:04saratogaBBB: ok i've got a few ffmpeg optimizations on my todo list
20:50:09saratogai'll try and get to them eventually
20:50:47saratogaTheSeven: did you enable USB disk mode for the nano2g yet?
20:51:35***Saving seen data "./dancer.seen"
20:52:53 Quit jfc^3 (Read error: 131 (Connection reset by peer))
20:54:50CIA-85New commit by Ubuntuxer (r23332): FS #10497 - New game codebuster, which is a clone of the classic game mastermind. It just runs on color LCD. Thank you to the author Clément ...
20:55:06TheSevensaratoga: I can't remember enabling it explicitly, but it seems to be enabled in SVN
20:55:20saratogaok was confused about people talking about it in the forums
20:55:23TheSevenmight have been done by someone else or sneaked in somehow
20:55:33TheSevenlet me blame it
20:56:07 Join Ubuntuxer [0] (n=johannes@
20:56:27 Quit funman ("free(random());")
20:56:42linuxstbUbuntuxer: Why the change to brickmania.tex in that commit?
20:56:54Ubuntuxerjust the id
20:57:00TheSevensaratoga: eek, it sneaked in in r23308 as a unrelated change
20:57:32BBBsaratoga, cool, thnx
20:57:43linuxstbUbuntuxer: Yes, but why as part of a completely unrelated commit?
20:57:44Ubuntuxeractually I want to make seperately
20:58:55saratogaany idea why USB HID makes a difference here?
20:58:57 Quit JackWinter (Read error: 104 (Connection reset by peer))
21:00:31 Join JackWinter [0] (n=jack@
21:04:32 Quit flydutch ("/* empty */")
21:08:12 Quit StealthyXIIGer (Read error: 104 (Connection reset by peer))
21:08:31 Join StealthyXIIGer [0] (
21:08:54 Join MethoS- [0] (n=clemens@
21:10:19TheSevensaratoga: I can't imagine what's causing that. But what annoys me far more than the HID stuff is that the FTL seems to be bugging again.
21:10:38saratogaok so long as you're aware that someone is having trouble with it
21:16:29saratogafunman: MT said he got a Clip+, any idea how difficult it will be to get a bootloader running on it?
21:16:34 Quit Ubuntuxer ("Leaving.")
21:19:01Tornelinuxstb: decryption failed, checksum error. but at least it doesn't die horribly.
21:20:38 Join phanboy4 [0] (
21:23:10linuxstbTorne: Yes, I've got an image of a 5.5g I can test with, and it dies with the same error. (and it's trying the same key)
21:23:54Torneonly the 3g aupd seems to ahve the flash image table in it
21:24:13Tornethe others don't have the hslf-type strings in at all
21:24:20Tornethough they are decrypted fine because i can disassemble them
21:24:34Torneso it looks like this might be excessively annoying regardless.
21:29:33CIA-85New commit by alex (r23333): Improve English in the codebuster manual entry.
21:29:45AlexPNow to correct spelling/punctuation in the plugin itself :)
21:29:50 Quit MethoS- (Remote closed the connection)
21:30:03 Join MethoS- [0] (n=clemens@
21:32:10 Quit BBB ("This computer has gone to sleep")
21:33:29*Dhraakellian updates
21:34:49linuxstbDhraakellian: No, it's nothing like mastermind. Completely different... No relation at all...
21:35:01Dhraakelliansorry, clone of mastermind
21:35:12linuxstbNo, it's nothing like mastermind. Completely different... No relation at all...
21:35:30DhraakellianI'll just have to see what the differences are, then. :D
21:40:46CIA-85New commit by alex (r23334): Correct spelling and punctuation in Codebuster menus.
21:40:48 Join Strife1989 [0] (
21:42:07saratogafunman: it was a data abort with the light off
21:43:07 Quit MethoS- (Remote closed the connection)
21:45:28saratogahuh somewhere in buffering, but nothing thats in the .map file
21:46:35 Quit Paprica ()
21:47:21 Join MethoS- [0] (n=clemens@
21:48:22 Quit MethoS- (Remote closed the connection)
21:49:01 Quit Strife89 (Connection timed out)
21:54:25AlexPWhat is going on with the Gigabeat S red in the current build?
21:54:40AlexPI don't think my change caused that :)
21:55:34 Join pamaury [0] (
22:00:03 Quit Buschel ()
22:00:49 Join BBB [0] (n=rbultje@
22:08:45*bertrik sees something resembling a rockbox logo on his meizu m6 sp \o/
22:09:11GeekShado_bertrik, awesome :)
22:11:03AlexPbertrik: Congrats :)
22:12:02bertrikit doesn't need a 24-bit framebuffer as I originally thought, it can work with 16-bit (RGB565) too
22:13:26bertrikGeekShado_, AlexP, n1s, gevaerts, thanks for the support! :P BTW, seeing a correct logo now (even the colors are right) :P
22:13:55GeekShado_sweet :)
22:14:27kugelbertrik: cool
22:14:39GeekShado_I hope your work willl work on M3 hardware too
22:15:22bertrikGeekShado_, the meizu m3 already has a working display, the m3 probably benefits most from the things that TheSeven is currently doing
22:15:38GeekShado_oh cool
22:17:34TheSevenmy nano has a current leak o.O
22:18:06AlexPTheSeven: At least it isn't a currant leak :)
22:18:23TheSevenas soon as i disable the USB controller once (and disable it again), power consumption more than doubles
22:18:44TheSevenI checked the USB powerdown code, and it seems to be right (and clocks are properly getting masked etc.)
22:19:34bertrikit's a really useful feature to have an adc for measuring current consumption :)
22:19:48 Part j-b
22:20:05TheSevenmeasuring it with a multimeter would in fact be better, but I can't reach the battery in that thing without destroying it
22:20:52TheSevenbut yes, this *would* be pretty useful for things like battery runtime estimation code, if rockbox would support it :-/
22:21:36kugelyou can make rockbox support it! :)
22:21:56bertrikoh I thought rockbox already did have a battery_current function or something similar
22:22:12TheSevenyes, but it's quite hardwired to
22:22:12bertrik(and using it for battery estimation)
22:22:21TheSeven*estimating* the current rather than measuring it
22:22:43kugelI think the estimation is actually taking the current into account, it's just that most/all assume a (compile-time) constant value
22:23:50n1slikely many older targets couldn't measure the current so rockbox never got the ability to support esimation based on measurment
22:24:29kugelthat's a reasonable assumption ;p
22:28:02bertrikthe meizu m6 sp display is currently in portrait mode, rather than landscape as it's used in the OF. I hope to find out more about the LCD to change this.
22:28:30 Part BBB ("Leaving")
22:28:33GeekShado_I remember I got a question
22:28:43GeekShado_in which langage are coded RockBox plugins ?
22:29:03TheSeven...or LUA
22:29:27TheSeven(whatever that is)
22:29:58n1syeah, i don't think any pugins have been coded in lua yet but it's possible
22:30:29TheSevenn1s: afair there is that viewport test plugin :-)
22:31:27n1sah, ok, that sounds useful :)
22:31:42CIA-85New commit by theseven (r23335): Plumb an iPod Nano 2G battery leak that started spilling as soon as USB was used at least once until a reboot.
22:32:26n1sReal Men (tm) use C anyway ;)
22:34:08n1sor RealvWomen
22:34:12n1s i suppose
22:34:20n1se /s/v/ /
22:38:46 Join Davide [0] (
22:40:02 Part Davide
22:40:06 Join Davide [0] (
22:40:10 Quit Davide (Client Quit)
22:40:21 Join Davide-NYC [0] (
22:41:19Davide-NYCHello all. I'm having big problems with a recent RB upgrade.
22:41:40saratogaTheSeven: should be farely easy to make runtime estimation use the current meter on the 2G
22:42:05saratogabut i wonder how useful it'll really be, at least with a compile time value we know the value chosen is an average over an entire battery bench
22:42:05Davide-NYCOn my H120-RTC-CFMod target I get codec and encoder errors as well as incompatible version. I've done clean installs (deleting .rockbox prior to unpack)
22:42:11Davide-NYCany ideas?
22:42:23saratogai think if you watch it in real time you may get a lot of jumping around as the bit rate changes and such
22:42:53saratogathough i suppose with a long enough running average it might work pretty well
22:42:55Davide-NYCCompiling in cygwin. reproduced problem on two different (although similarly modded) units.
22:46:37TheSevenhow can modifying system_exception_wait make it crash on boot, reproducibly?
22:48:20TheSevenok, what should i fix/break next? :-)
22:50:47n1sDavide-NYC: you don't runa rockbox from flash, do you?
22:50:47*TheSeven is about to file a flyspray bug regarding flyspray being unable to search for all bugs affecting a certain target
22:50:48Davide-NYCDoh! I think I figured it out. I forgot to flash my RAM/ROM
22:50:53n1srun, even
22:51:25Davide-NYCn1s: exactly my prblem, but I just formatted my CF and am compiling my code (again) so in a minute we'll see.
22:51:39***Saving seen data "./dancer.seen"
22:51:48saratogarunning from flash if you have a CF card probably isn't worth the hassel
22:51:50n1sthat is likely the problem
22:51:54Davide-NYCI just forgot the flashing step so everything on the CF itself was truly incompatible
22:52:54Davide-NYCsaratoga: I'm not sure booting from CF is supported by bootloader v7pre4
22:54:14n1sDavide-NYC: there is some kind of way to boot from the hd with that bootloader that you need to use anyway if you want to flash IIRC
22:54:37n1sbut i think that is new enough to have the cf fixes
22:55:44Davide-NYCany news on a final V7 release for that bootloader?
22:56:49*TheSeven wonders if it's worth a try to dig into all that UI stuff to track down what's going on with nano2g scrolling
22:57:09 Join bluebrother [0] (
22:58:39 Quit bluebroth3r (Read error: 60 (Operation timed out))
22:59:48 Quit pamaury ("exit(*(int *)0 / 0);")
23:00:08Davide-NYCsaratoga: booting from CF (HD) works perfectly.
23:03:10saratogaanyone feel like trying out a clip build and letting me know the abort address when it crashes?
23:04:54*amiconn returns
23:05:43amiconnkugel: At the time you linked to you only said that you experimented a bit, but there are no results given
23:05:58amiconnDid you measure the potential speedup meanwhile?
23:06:23kugelamiconn: well, I asked you if you saw it. if you said no I would've shown you some results
23:06:33*amiconn also wonders whether it'd matter in real-world figures at all
23:07:27bertrikamiconn, I'm reading a datasheet right now that indicates that it should indeed be possible to change the portrait/landscape mode for some targets, I'll try that now (bit 3 in R03h - Entry mode)
23:07:33kugel - basically 10% speedup on my fuze, <5% on my e200
23:07:49kugelthe results are further down
23:08:09kugelstarting at 2:45
23:10:46amiconnI still wonder how important it is. The usb screen is one of the rare cases that uses transparent bitmap drawing for a large bitmap, isn't it?
23:11:00amiconnAlso, is the asm function smaller or larger than the C equivalent?
23:12:52kugelI didn't messure binsize if you mean that
23:13:11bertrikaw, the landscape/portrait swap didn't work ..
23:13:36amiconnI do mean that
23:13:57amiconnThe place where this function resides currently is odd.
23:14:03 Quit polobricolo (Read error: 110 (Connection timed out))
23:14:05kugelthe inline assembly is only 15 lines, it doesn't make a difference anyway
23:14:28kugelif we have it already we can use it for all (applicable) targets IMO
23:15:09 Quit tomers (Read error: 60 (Operation timed out))
23:15:10*amiconn thinks this could be sped up even further, at least on ARMv6 (beast)
23:15:52 Join dockimble [0] (n=dockimbl@
23:17:28 Quit n1s ("Lmnar")
23:17:35kugelit has a tad bit different clipping and doesn't calculate dst_end, I wonder what code to keep (apart from the asm)
23:19:00 Join faemir [0] (n=faemir@
23:21:01amiconnThe C code needs to stay anyway, for the sims and the other archs
23:21:27kugelI mean the code outside the loop
23:24:19Davide-NYCComment about iRiver bootloader v7pre4: On CFModded H1x0 targets USB disk mode does not work for all CFCards. Neither does booting from CFCard (for probably the same as of yet unknown reason).
23:24:56amiconnThe C code isn't very efficient
23:25:33*amiconn guesses that if the inner loop in the C code is optimised, it will be as fast as the asm, with the added benefit that it'd also speed up things on the other archs
23:27:07amiconnIn fact I guess that if this gets optimised, the compiled arm code will look very similar, if not identical, to the asm version
23:29:09kugelthat wouldn't surprise me
23:29:26 Join dmb [0] (n=Dmb@unaffiliated/dmb)
23:29:41 Quit dmb (Read error: 104 (Connection reset by peer))
23:31:33 Join dmb [0] (n=Dmb@unaffiliated/dmb)
23:32:01 Join reydarz [0] (
23:33:17kugelamiconn: so what now?
23:33:55amiconnImo optimising the C would be the best solution
23:34:31amiconnLess ifdef,s and more targets benefit from that
23:35:52kugelI don't think I'm able to optimize it to be as fast, I could try the count-towards zero thing (but that might actually be sufficient already) but I don't see much more
23:36:35amiconnThe inner loop is a for loop atm. That's not the best choice speed wise
23:36:42Davide-NYCI just compiled a patched bootloader from FS #9642, I'm expecting a file called "bootloader.iriver" but instead I have a bootloader.bin file. Are these the same file?
23:36:48*amiconn might have a shot at it tomorrow
23:37:35Davide-NYCGuys, I'm going to be flashing my bootloader and do not want to brink my beloved H1120+mods. Am I doing this correctly?
23:38:23AlexPI wouldn't be flashing an untested H100 bootloader if it were me
23:38:44Davide-NYCAlexP: can you take a look at FS #9642?
23:39:15Davide-NYCI mean, I've gotten a lot of good use out of RB and *somebody* has to test this patch.
23:39:38AlexPDavide-NYC: I'm not testing any H100 bootloader patches
23:39:53AlexPDavide-NYC: I have no way of recovering if it is not good
23:40:25*TheSeven wonders how that bootloader was developed then
23:40:27saratogasomeone with a BDM cable should test it so that they can unbrick if needed
23:40:51AlexPTheSeven: By linusn who has a BDM wiggler
23:40:51Davide-NYCNo, I'm willing to do it, since I have a spare unit. I was just asking that you look at the patch since I'm not very good at coding and there is a persistent bug in the BL as is.
23:41:08AlexPDavide-NYC: I'm not the person to be commenting on the code :)
23:41:36AlexPI could, but it'd be meaningless :)
23:42:37Davide-NYCQuestion remains: Is bootloader.bin == bootloader.iriver? (presuming I've configured correctly?
23:43:08AlexPTheSeven: The H100 and H300 bootloaders are a bit of a bugger - they need thorough testing by someone who can recover in the event of problems before release, and this person is LinusN (who is busy and not around much these days). That means that they don't tend to get updated
23:43:41AlexPDavide-NYC: One mo
23:46:47 Quit dmb ("Leaving")
23:48:52AlexPDavide-NYC: Sorry, not sure. I was looking something up but couldn't find it
23:49:39Davide-NYCme neither :(
23:51:42Davide-NYCI just tried to flash the bootloader.bin file and the iriverflash utility gave an "unkown filetype" error.
23:51:48TheSevenguessing from the name, it may need to be run though scramble, but honestly I have no idea about that target
23:51:54 Join mt2 [0] (n=mtee@
23:52:13 Nick mt2 is now known as mt (n=mtee@
23:53:11 Join dmb [0] (n=Dmb@unaffiliated/dmb)
23:53:56Davide-NYCpetur: ping
23:54:24SlasheriDavide-NYC: NO, bootloader.bin is not bootloader.iriver
23:54:33Slasherirockbox.iriver is the bootloader.iriver
23:54:39Slasheri(when compiling bootloader)
23:54:41Davide-NYCHow do I "make" that?
23:54:54Slasheriyou should already have it, if you compiled the bootloader
23:55:09saratogahow trustworthy are data abort addresses on AMS?
23:55:35Slasheribut in order to flash it, you need to override safety mechanism from iriver_flash.c utility (or add then checksum of the new file there)
23:55:35Davide-NYCSlasheri: Got it.
23:55:43*TheSeven tries to get grip on the GUI code and terribly fails...
23:56:34saratogaif i have the address can I disassemble and find the exact line or is it only approximate?

Previous day | Next day