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).

Notice: Only Gecko based browsers prior to FF4 support the multipart/mixed "server push" method used by this log reader to auto-update. Since you do not appear to use such a browser, this page will simply show the current log, and not automatically update.

#rockbox log for 2010-01-03

00:07:30 Quit togetic (Read error: 110 (Connection timed out))
00:09:02 Join togetic [0] (n=togetic@unaffiliated/ibuffy)
00:14:10 Join Mitsukaru [0] (i=4c5b0124@gateway/web/freenode/x-lyyqccgwwixxbcen)
00:14:27***Saving seen data "./dancer.seen"
00:22:25Mitsukarusomeone asked me if they could install rockbox on a gigaware player. could it work, even if it isnt
00:22:31 Quit Bagder ("It is time to say moo")
00:22:37Mitsukarueven if it isnt "supported"
00:23:06Mitsukaruquote button is right nex to the return key, argh >:/
00:23:19mc2739Mitsukaru: no. Rockbox is tailored to each platform
00:23:45Mitsukaruoh. so there's no way it could work then?
00:24:12Unhelpfulportions of it must be rewritten for each new player. and "new player" might mean a device that looks the same and is called by the same name as one from last year.
00:24:58 Quit grndslm ("Leaving")
00:25:23mc2739only if somebody does the work of porting to that player
00:25:48mc2739Mitsukaru: only if somebody does the work of porting to that player
00:25:51Mitsukaruhow easy is it to port it to a player?
00:26:08JdGordon1not
00:26:24mc2739Mitsukaru: http://www.rockbox.org/wiki/NewPort
00:26:27Mitsukaruah, i see. thanks
00:26:32Unhelpfulit is not a trivial job and pretty much *has to* be done by somebody who has one.
00:29:30 Join Barahir [0] (n=jonathan@gssn-5f755148.pool.mediaWays.net)
00:31:01 Quit togetic (Read error: 110 (Connection timed out))
00:32:16 Quit Mitsukaru ("thanks")
00:38:16 Quit markun (niven.freenode.net irc.freenode.net)
00:38:16NSplitniven.freenode.net irc.freenode.net
00:38:16 Quit AndyI (niven.freenode.net irc.freenode.net)
00:38:16 Quit parafin (niven.freenode.net irc.freenode.net)
00:38:16 Quit ps-auxw (niven.freenode.net irc.freenode.net)
00:38:16 Quit bzed (niven.freenode.net irc.freenode.net)
00:38:16 Quit jon-kha (niven.freenode.net irc.freenode.net)
00:38:16 Quit toffe82 (Read error: 60 (Operation timed out))
00:40:15 Join kugel_ [0] (n=kugel@e178083061.adsl.alicedsl.de)
00:40:34 Quit bluebrother ("Get MacIrssi - http://www.sysctl.co.uk/projects/macirssi/")
00:41:15 Quit merbanan (Read error: 110 (Connection timed out))
00:42:35NHealniven.freenode.net irc.freenode.net
00:42:35NJoinAndyI [0] (n=pasha_in@212.14.205.32)
00:42:35NJoinmarkun [50] (n=markun@rockbox/developer/markun)
00:42:38NJoinbzed [0] (n=bzed@devel.recluse.de)
00:42:38NJoinjon-kha [0] (i=jon-kha@kahvi.eu.org)
00:42:38NJoinps-auxw [0] (n=arneb@dyn37.ps-auxw.de)
00:42:38NJoinparafin [0] (i=parafin@paraf.in)
00:55:01 Join perfectdrug [0] (n=5b0ee037@giant.haxx.se)
00:55:38perfectdrugpixelma: ping:)
00:56:13CIA-6New commit by 03bertrik (r24150): Make debug builds compile again
00:58:40 Quit kugel (Read error: 110 (Connection timed out))
00:58:52 Quit JdGordon1 (Read error: 110 (Connection timed out))
01:00
01:02:04 Part mahoney
01:03:59soapsaratoga, When you say "with current versions of foobar" - how current are we talking?
01:04:21saratogasoap: I'm on a 1.0 beta, but thats been there for a long time i think
01:04:34saratogai thought it was added early in the 0.9 series
01:05:16soapyea, it's there in my version.
01:05:31soapI'll make a test file and see if that forces a rewrite or not.
01:05:39saratogaIIRC before it was added you had to change the default tag type, check that option you mentioned, and then force a tag rewrite through the context menu
01:05:47soapIF his problem is APE RG tags, though, my first test says that will do it.
01:05:59saratogasomething like "rewrite tags from database" that no longer exists
01:06:39soapso I guess I'm drifting more OT in wondering if it would have fixed my bastardized mix of 2.3 and 2.4 tags.
01:07:02soapback OT, I still like my APE RG tag theory, though.
01:07:34saratogaiirc you can tell foobar to write standard v2.4 tags, or backwards compatible but less standard 2.3 tags
01:07:41saratogaboth should work with rockbox
01:08:20saratogafwiw i still think we should have just accepted the apev2 tagging patch and avoided this potential confusion
01:08:21soapright, not labeled as such, rather labeled "compatibility mode" or not.
01:08:49saratogayeah you have to check the faq to get the details
01:13:17saratogaFP emulation is allowed in plugins right?
01:16:30kugel_everything is allowed there :)
01:17:08kugel_well, calculator uses fp emulation
01:17:28 Join grndslm [0] (n=grndslm@174-126-14-4.cpe.cableone.net)
01:23:01pixelmaperfectdrug: hi :) wanted to ask you about http://www.rockbox.org/tracker/task/10412... do you use a WPS with albumart (and have pictures around)?
01:23:41pixelmareason: I can reproduce on my M5 only with an AA WPS
01:25:34 Join Kopfgeldjaeger [0] (n=nicolai@monitor-mode-enabled-on-mon0.phy0.de)
01:25:52pixelmaand pictures present
01:27:01 Join S_a_i_n_t [0] (i=S_a_i_n_@203.184.3.82)
01:27:13funmani'm looking at merging fuze & e200v2 lcd drivers, perhaps the difference in scrollwheel is there
01:27:31perfectdrughi pixelma, what do you mean with "have pictures around", I don't have an iaudio myself I just reported this bug for someone else without a flyspray account and prober english knowledge, but I can ask him, I guess though he uses the standard cabbie theme.
01:28:10perfectdrugah i see what you mean
01:29:00perfectdrugyou think it could be a problem with the jpeg AAloader or something
01:29:06pixelmaperfectdrug: well an album art WPS (cabbie would be) and also the bitmap files so that the art will be displayed
01:30:55 Quit bertrik ("De groeten")
01:31:39 Quit GodEater (Read error: 113 (No route to host))
01:33:24perfectdrugI ask him to try with a non album-art wps and report back, anything else to test? I offered him to track down the problem to a specific revision but he said he just replaced the hdd and the battery and wanted to do that some time later maybe.
01:34:37 Quit domonoky (Read error: 104 (Connection reset by peer))
01:36:16 Join GodEater [0] (n=bibble@bb-87-80-121-64.ukonline.co.uk)
01:37:57 Join Strife89 [0] (n=michael@adsl-146-208-54.mcn.bellsouth.net)
01:40:48pixelmaI don't think so, just wanted some more data, I could reproduce reliably after finding that out (amiconn could not with his X5 - using a non-AA WPS) so I could try tracking it down myself. Looking at the date of the bugreport it seems to be there quite a while :\
01:42:34pixelmaan interesting test might be if it is reproducible on a coldfire Iriver to see if it depends on CPU architecture (as I couldn't reproduce on my c200) and there are a few other differences between Iaudios and those Irivers
01:46:01saratogakugel_: do you have any opinion about my clip poweroff patch?
01:46:27kugel_?
01:46:32kugel_which patch?
01:47:18saratogaFS #10887
01:49:30saratogaits a minor ui annoyance i'd like to fix, but i'm not sure if theres parts of rockbox (plugins?) that depend on being able hold power and some other button at the same time without shutting down
01:49:34perfectdrugpixelma: ok I told him to test and I update the flyspray task with his findings, in the meantime I can try to reproduce it on my h340, but the iriver has its own stopbutton so the long pause doesn't work anyway
01:49:49kugel_saratoga: it affects all targets as far as I can see it?
01:49:53pixelmapr
01:49:56saratogakugel_: yes
01:50:20saratogai can ifdef it if needed, but its annoying to me on several targets so i didn't see much sense in ifdef'ing it
01:50:37pixelmaperfectdrug: I don't think the buttons have anything to do with it, just that music is completely stopped
01:51:13pixelmaand thanks for testing :)
01:52:59funmansaratoga: i think it's ok
01:54:20saratogaok cool
01:54:31saratogathink i should ask on the dev list to be sure no one dislikes the idea?
01:54:32Unhelpfulthere's also a ~40cycle arm9e and up divider in the ARM system developer's manual, but it uses a lot more asm tricks and would be much harder to reimplement without copying outright... which it really doesn't *look* like the book permits me to do.
01:54:34 Join toffe82 [0] (n=chatzill@adsl-75-3-221-212.dsl.frs2ca.sbcglobal.net)
01:55:10perfectdrugpixelma: I have EQ and bass applied all the time, and never experienced something like this at all over the years, I also wonder why such a bug does not affect and annoy more people, but could be that the iaudio is not such a popular rockbox device.
01:55:23saratogaalso, does anyone understand what the "running time" menu is supposed to do?
01:55:30Unhelpfulit would be much nicer than using the VFP on the beast, though, and would even benefit a few other targets.
01:55:38 Part xnyhps
01:55:40saratogalooking at the code it seems to count how much time has passed since the last time someone looked at it while charging
01:55:44gevaertssaratoga: probably a stupid question, but could FS #10887 interfere with hold in any way?
01:55:57saratogagevaerts: I'm not sure
01:55:59saratogalet me try
01:56:13saratogawell it can't on the clip since hold and power are the same slider :)
01:56:14kugel_saratoga: I think it could hurt on a number of targets
01:56:33kugel_esp. those with little buttons (ondio, ipods)
01:57:43 Quit Omlet ("( www.nnscript.com :: NoNameScript 4.22 :: www.esnation.com )")
01:58:30saratogakugel_: so there are targets that have long power + some other button defined?
01:59:15Unhelpfulsomebody else want to rewrite it in asm after i write it in C? ;)
01:59:37perfectdrugpixelma: my theory is that because the play button is mapped to 3 different functions on the iaudio (AFAIK), that rockbox somehow confuses stop (long press) with shutdown (longer press) and shuts down the audio or something but not the device, but that is just a wild guess and completly unlogical with your AAtheory
02:00
02:00:08pixelmaperfectdrug: how likely it is that you experience the bug depends on your listening habits. If you just turn the device on and resume or start a new playlist you won't hear it
02:00:40pixelmapower off is on a seperate "button", not a longer press of play
02:00:50perfectdrugon the X5?
02:00:55pixelmayes
02:01:32perfectdrugoh so I may confuse this with the remote he was talking about
02:01:40pixelmaM5 and X5 have a power button which is used to switch off the device
02:01:56kugel_saratoga: I think so, yes
02:02:17pixelmacould be, I don't have a remote so don't know the mappings
02:02:48saratogagevaerts: on the sansa at least hold overrules power even with these changes, but i dont' know if thats hardware dependent
02:03:01perfectdrugpixelma: as he said he can reproduce it without a remote I try to reproduce it on my iriver
02:03:12saratogaactually i don't think it is since the sansa bootloaders for the longest time would ignore hold until we fixed them
02:03:41gevaertssaratoga: ok, so that's not an obstacle anyway
02:04:10saratogaimo unless there are targets that absolutely need power as a combo, holding power should always shut down the player
02:04:19 Quit GodEater (Read error: 113 (No route to host))
02:05:10 Join Sajber^1 [0] (n=Sajber@h-142-12.A213.priv.bahnhof.se)
02:05:25 Quit Sajber^ (Read error: 110 (Connection timed out))
02:05:25gevaertsI tend to agree
02:05:31*perfectdrug thinks that it is a shame that my 60gig hdd is dying slowly in my iriver:/
02:05:52 Join GodEater [0] (n=bibble@bb-87-80-121-64.ukonline.co.uk)
02:06:11funmansaratoga: no buttons should be returned at all if hold is set
02:06:51kugel_saratoga: I think even the e200 has a combo with power, dir skip, I can imagine such combos are used for not so frequent actions like a-b-repeat or something similar
02:07:15kugel_and you don't really have a choice on some targets
02:07:42saratogawhich screen is dir skip?
02:08:14saratogaoh i see it
02:08:32saratoga ACTION_WPS_ABSETB_NEXTDIR, BUTTON_POWER|BUTTON_RIGHT, BUTTON_POWER
02:08:47saratogawhat does the BUTTON_POWER, BUTTON_POWER option mean?
02:08:52saratogathat you just hold power ?
02:09:35saratogawow thats a neat feature, i didn't realize we had it
02:09:35 Quit funman ("free(random());")
02:09:49saratogabut even with this patch it works fine, as long as you don't hold power for a couple seconds
02:11:17pixelmait checks if you are holding the power button before the right button, so you don't need to press the keys at *exactly* the same time
02:11:28saratogaah ok
02:13:30saratogawhat is button_context_tree ?
02:14:28***Saving seen data "./dancer.seen"
02:14:37pixelmathe button action context for the tree (filebrowser etc. I think)
02:15:05pixelmaoff the top of my head
02:15:57saratogapixelma: so thats the context menu in the file browser?
02:16:08pixelmano
02:17:04pixelmathe button context is the set of mappings that belong to a "screen" so to speak
02:17:31saratogaso how would i get to a place where i could use them?
02:17:43pixelmahmm... that explanation sounds a bit weird
02:17:58kugel_I think fs10883 has found the root problem that sometimes keys do not work on scrollwheel targets
02:18:07kugel_after exiting some plugins for example
02:18:33pixelmasaratoga: could you rephrase?
02:19:06saratogapixelma: i want to use one of the buttons on the screen, how do i get to a place where rockbox will let me do that
02:19:18saratogae.g. how do I generate a "ACTION_TREE_STOP"
02:19:39kugel_the context is a parameter of get_action()
02:20:07saratogaits the only thing in a supported rockbox port that has BUTTON_POWER|BUTTON_REPEAT mapped
02:20:08kugel_i.e. whereever you call get_action with CONTEXT_TREE, it will generate ACTION_TREE_STOP if you press power
02:20:57saratogawhen would get_action be called? i've never looked at this part of the code
02:21:08pixelmasaratoga: press the button that are mapped to ACTION_TREE_STOP in e.g. the filebrowser while music is playing
02:21:25kugel_in the main loop of a screen, where the button handling (i.e. waiting for user input) happens
02:21:25saratogaoh
02:22:06saratogaso on the gigabeat F you hold power to do that?
02:22:14saratogai'm pretty sure that just shuts down the F
02:22:23saratoga{ ACTION_TREE_STOP, BUTTON_POWER|BUTTON_REPEAT, BUTTON_NONE },
02:22:35saratogaor does the BUTTON_NONE make that not happen?
02:25:54kugel_button_none there makes it not check for a specific button before the actual combo
02:26:15pixelmano - the last parameter in the list is the pre condition so check if a certain button is pressed before the other (for combos or distinguishing long and short presses). BUTTON_NONE just means that it isn't specified and any button could be held before. I don't know what would happen an a Gigabeat if you hold power longer. don't have one
02:26:29kugel_ ACTION_WPS_ABSETB_NEXTDIR, BUTTON_POWER|BUTTON_RIGHT, BUTTON_POWER means that BUTTON_POWER needs to be pressed before the combo is evaluated
02:26:49saratogakugel_: so if the combo is BUTTON_POWER|BUTTON_REPEAT but it doesn't check for a previous button, what is BUTTON_REPEAT doing?
02:27:10saratogadoes the repeat override the NONE?
02:27:22kugel_any button is OR'd with _REPEAT after some ticks
02:27:43kugel_if you release it it's OR'd with BUTTON_REL
02:28:45pixelmaperfectdrug: I need some sleep now (but stay connected), you could post any findings here or in the tracker. Thanks again
02:29:30saratogakugel_: now i'm confused
02:29:35saratogaon the F, its ACTION_TREE_STOP, BUTTON_POWER|BUTTON_REPEAT, BUTTON_NONE
02:29:48saratogabut if i tap the POWER as quickly as I can it stops
02:29:53saratogaor it stops if I hold it
02:30:10kugel_if you tap it only quickly it stops?
02:30:16saratogayeah
02:30:25kugel_that shouldn't happen
02:30:55saratogai tapped it as fast as I possibly could
02:31:15saratogaand if I hold it it still stops, and if i keep holding it, then the F stops playback then a second later shuts down
02:31:34kugel_saratoga: well, there are multiple entries for ACTION_TREE_STOP, that's why
02:31:46saratogabah i just looked at the line returned by grep
02:39:01saratogai've noticed some plugins disable powering off via POWER, how exactly does that work? do they just swallow the button event?
02:42:16saratogais AB mode in the manual? i can't seem to find it
02:42:38saratoganevermind
02:45:07saratogaso i've enabled AB mode but i can't seem to actually get it to work on the e200
02:46:07saratogaoh hm it works, its just really laggy so its hard to notice
02:52:28saratogakugel: ok after a lot of grepping I think the only place that maps POWER to a button combo is ABREPEAT on the sansas and sa9200
02:52:41saratogaand the H10 which has a combo for ACTION_WPS_VIEW_PLAYLIST
02:53:38 Part S_a_i_n_t
02:53:42saratogaand for whats its worth, the power off time is longer then the other times, so you can still use AB repeat pretty easily
02:53:51saratogaat least on my sansa, i don't have an H10
03:00
03:02:42 Quit Sajber^1 ("Leaving.")
03:05:56 Join S_a_i_n_t [0] (i=S_a_i_n_@203.184.3.82)
03:06:16 Quit kugel_ ("exit(0);")
03:10:18 Join JdGordon| [0] (n=me@rockbox/developer/JdGordon)
03:12:01 Quit JdGordon| (Remote closed the connection)
03:14:24 Quit froggyman (Read error: 104 (Connection reset by peer))
03:15:12 Join froggyman [0] (n=sopgenor@pool-72-69-220-194.chi01.dsl-w.verizon.net)
03:28:11 Quit Unhelpful (Read error: 104 (Connection reset by peer))
03:28:24 Join Unhelpful [0] (n=quassel@pool-71-173-205-32.hrbgpa.fios.verizon.net)
03:28:46 Quit Unhelpful (Remote closed the connection)
03:28:59 Join Unhelpful [0] (n=quassel@pool-71-173-205-32.hrbgpa.fios.verizon.net)
03:33:34 Join kramer3d [0] (n=kramer@unaffiliated/kramer3d)
03:34:26 Quit efyx (Remote closed the connection)
03:46:16Unhelpfulsaratoga: fyi we definitely cannot reduce udiv accuracy to speed up APE, udiv happens in the bitstream decoder and the codec *will* crash on wrong results, as i've been spending all day learning.
03:46:45saratogaflow control depends on the result of an FP divide?
03:46:55saratogain a lossless codec
03:48:09Unhelpfulno, an integer divide.
03:48:35saratogaah ok then
03:49:55Unhelpfuland we're down to 81.04MHz :)
03:51:13Unhelpfuli could shave another .05s by not trapping divide-by-zero... it's probably not worthwhile.
03:54:12saratogahave you profiled the codec?
03:55:32 Quit Tomis ()
03:55:33Unhelpfulsaratoga: i haven't, but i know how many divides it does while decoding the test file, because i logged their ranges on the sim - the new divider is optimized for numerators with the high bit unset, about 3 or 4 divides of about 4.5 million done while decoding the test file do not meet this requirement.
03:56:49saratogaUnhelpful: so much time does the divide use?
03:57:21saratogaor better yet how many MHz
04:00
04:00:43Unhelpfuli don't know precisely, but the fact that i've only saved about 5-10 cycles with today's work and brought us down a MHz or so suggest quite a bit :)
04:02:03saratogaUnhelpful: how many calls per second?
04:02:35saratogaoh 3*88200
04:03:26saratogathat would imply 4 cycles saved if you got a 1MHz speedup
04:07:00saratogaunless you only call it once per stereo channel pair, in which case 8 clocks
04:10:43 Quit GodEater (Read error: 60 (Operation timed out))
04:12:16 Join GodEater [0] (n=bibble@bb-87-80-121-64.ukonline.co.uk)
04:14:30***Saving seen data "./dancer.seen"
04:15:26 Quit Strife89 ("Bed.")
04:17:47 Part froggyman
04:25:54 Quit GodEater (Read error: 60 (Operation timed out))
04:27:22 Join Barahir_ [0] (n=jonathan@gssn-5f757704.pool.mediaWays.net)
04:28:03 Join GodEater [0] (n=bibble@bb-87-80-121-64.ukonline.co.uk)
04:28:28Unhelpfulsaratoga: amiconn said it should be 3 per sample - i assume that's for stereo. the actual savings is a bit complicated, since a lot of it came from removing short tests for special cases. there's a gain from not doing the test on every divide, but also a loss from the special case being done by the main divide loop, which takes somewhere around 25 + result bits * 2 cycles
04:29:27Unhelpfulthis works out nicely for the divisor-exceeds-numerator special case, as there's only one pass done of the main divide loop. for the divisor-is-one and divisor-is-power-of-two special cases the saving varies based on numerator size.
04:29:37 Join JdGordon1 [0] (n=jonno@m365736d0.tmodns.net)
04:30:49Unhelpfulthe total number of loop passes is numerator bits - divisor bits + 1, and the loop itself is 2 cycles per pass, unless there's a numerator with the high bit set. svn currently falls back to the old divider for this case, i'll commit shortly a version that shifts the input right on, does a 31-bit divide, and then fixes the result.
04:32:38 Quit DerPapst1 ("Leaving.")
04:33:36Unhelpfulshould probably also use this divider on armv5. armv6 targets are all arm11 cores, right? that takes 3 cycles/pass for the new loop, thought there may still be a small code size improvement
04:34:17 Quit GodEater (Remote closed the connection)
04:34:33 Join GodEater [0] (n=bibble@bb-87-80-121-64.ukonline.co.uk)
04:36:17Unhelpfulthe savings outside the divide loop on the special-case removal appears to be about 6 cycles... but of course there are 2 cycles used to test for the large numerator and for the skipped branch if the the test fails.
04:39:25 Quit Barahir (Read error: 110 (Connection timed out))
04:39:39 Quit Lss (Read error: 104 (Connection reset by peer))
04:40:15 Join togetic [0] (n=togetic@unaffiliated/ibuffy)
04:42:08 Join ajjsjs [0] (i=S_a_i_n_@203.184.0.57)
04:42:29 Quit S_a_i_n_t (Read error: 110 (Connection timed out))
04:42:52 Quit ajjsjs (Client Quit)
04:42:55 Join S_a_i_n_t [0] (i=S_a_i_n_@203.184.0.57)
04:43:20 Part S_a_i_n_t
04:43:32 Join S_a_i_n_t [0] (i=S_a_i_n_@203.184.0.57)
05:00
05:06:30 Join Tomis [0] (n=Tomis@70.134.65.110)
05:16:34 Quit saratoga ("Page closed")
05:18:47 Quit JdGordon1 (Read error: 104 (Connection reset by peer))
05:23:44 Join JdGordon1 [0] (n=jonno@m465736d0.tmodns.net)
05:30:15CIA-6New commit by 03unhelpful (r24151): More comments for udiv32_armv4.S, reduce zero divisor test to one cycle for the skipped branch by setting flags when inverting divisor, 32-bit ...
05:37:47 Join Sajber^ [0] (n=Sajber@h-142-12.A213.priv.bahnhof.se)
05:40:02 Join froggyman [0] (n=sopgenor@pool-72-69-220-194.chi01.dsl-w.verizon.net)
05:48:20CIA-6New commit by 03unhelpful (r24152): Use long jump to reach __div0 from udiv32_arm if building with IRAM and without EABI.
05:48:52 Quit Sajber^ (Read error: 104 (Connection reset by peer))
05:51:11 Quit CaptainKewl (Remote closed the connection)
06:00
06:14:34***Saving seen data "./dancer.seen"
06:20:30 Quit kadoban (Remote closed the connection)
06:22:20 Quit togetic (Read error: 110 (Connection timed out))
06:23:01 Join togetic [0] (n=togetic@unaffiliated/ibuffy)
06:25:51 Join saratoga [0] (i=9803c6dd@gateway/web/freenode/x-tfwlpruwdzayzumn)
06:26:13saratogai got tired of waiting for all those lazy ipod classic and nano4g owners to start a port, and made a wiki page for them
06:30:05 Quit JdGordon1 (Read error: 104 (Connection reset by peer))
06:43:45 Quit fdinel ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
06:44:46 Quit kramer3d (Read error: 104 (Connection reset by peer))
06:45:06 Join kramer3d [0] (n=kramer@unaffiliated/kramer3d)
06:48:53 Join getchar [0] (i=Aeromani@201.32.97.202)
06:57:57 Quit perfectdrug ("CGI:IRC")
07:00
07:11:57 Quit getchar (Client Quit)
07:13:43 Quit SolarWar (Read error: 110 (Connection timed out))
07:36:37 Join Tomis2 [0] (n=Tomis@70.134.77.113)
07:39:37 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
07:53:00 Quit Tomis (Read error: 110 (Connection timed out))
07:53:01 Nick Tomis2 is now known as Tomis (n=Tomis@70.134.77.113)
08:00
08:02:37 Quit bmbl (Excess Flood)
08:02:41 Part froggyman
08:03:03 Join AndyIL [0] (n=pasha_in@212.14.205.32)
08:09:05 Quit togetic (Read error: 110 (Connection timed out))
08:10:03 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
08:14:37***Saving seen data "./dancer.seen"
08:17:08 Quit AndyI (Read error: 110 (Connection timed out))
08:19:11 Quit bmbl (Excess Flood)
08:19:30 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
08:51:15 Quit toffe82 (Read error: 60 (Operation timed out))
09:00
09:24:27amiconnUnhelpful: There was a reason why I enabled the special division routine for ARMv4 only. For ARMv5 and higher, libgcc as an equally efficient routine (using clz of course), just the ARMv4 one was crap
09:24:47amiconnMaybe your improved 2 insns/cycle mechanism is better though
09:25:37amiconnsaratoga: The runtime menu displays the battery runtime since the device was last charged, and also remembers the max. runtime of of all discharge cycles
09:26:38amiconnBoth times are manually resettable, since on several targets rockbox can not/ not reliably/ not always detect whether the device is/was charged
09:41:46 Join flydutch [0] (n=flydutch@host51-165-dynamic.15-87-r.retail.telecomitalia.it)
09:52:03 Nick Barahir_ is now known as Barahir (n=jonathan@gssn-5f757704.pool.mediaWays.net)
10:00
10:01:06 Join stoffel [0] (n=quassel@p57B4D8AA.dip.t-dialin.net)
10:10:15 Join togetic [0] (n=togetic@unaffiliated/ibuffy)
10:12:14 Join Grahack [0] (n=Grahack@ip-222.net-82-216-222.rev.numericable.fr)
10:14:42***Saving seen data "./dancer.seen"
10:28:31 Join bertrik [0] (n=bertrik@ip117-49-211-87.adsl2.static.versatel.nl)
10:37:54 Join merbanan [0] (n=banan@c-83-233-52-241.cust.bredband2.com)
10:46:55 Quit kaniini (Read error: 110 (Connection timed out))
10:57:15Unhelpfulamiconn: if all of our ARMv5 targets are ARM9e cores (or better) with the faster multiplier and SMULWy, there's a newton-raphson divider from the ARM system developer's guide with performance of 27-36 cycles depending on divisor size. it'll be some work documenting the algorithm sufficiently to reimplement it, though - the book doesn't specify any separate copyright for code that i can find and has the usual sort of all rights reserved
10:57:15Unhelpful language regarding the text.
11:00
11:00:27Unhelpfullooking at firmware/export/config* the beast is pretty much the most-finished target that would benefit?
11:04:44 Join stooo [0] (n=sto@g226200059.adsl.alicedsl.de)
11:08:32Unhelpfulgcc devs seem to think that the special cases i took out are generally useful, too? libgcc for arm9e and arm926ej-s is doing essentially exactly what the old udiv32_arm was doing. on targets where there's not an interlock stall on early registers the new udiv should still be faster than the old, and the shift sequence for normalization can be replaced with two clzs and a bit of math to clamp the difference to the right range.
11:08:42stooohi
11:11:30Unhelpfulon arm11 that would only save code size, since the two-instruction divide loop body will interlock and stall to three instructions... although i wonder, if the subcc doesn't cause an interlock when skipped it might be 3 instructions per unset bit in result, 2 instructions per set bit in result.
11:18:54 Join pamaury [0] (n=pamaury@ALyon-551-1-5-153.w92-137.abo.wanadoo.fr)
11:19:45CIA-6New commit by 03Buschel (r24153): Fix tabs
11:27:44CIA-6New commit by 03Buschel (r24154): Fix more tabs
11:35:33CIA-6New commit by 03Buschel (r24155): Fix even more tabs
11:38:17 Quit stoffel (Read error: 60 (Operation timed out))
11:38:44 Join stoffel [0] (n=quassel@p57B4DB1A.dip.t-dialin.net)
11:38:53 Join Buschel [0] (n=ab@p54A3FB4D.dip.t-dialin.net)
11:41:13 Join DerPapst [0] (n=DerPapst@91-64-221-175-dynip.superkabel.de)
11:48:34 Quit antil33t ()
11:50:35CIA-6New commit by 03Buschel (r24156): Fix further tabs
11:54:45 Join antil33t [0] (n=Mudkips@203-184-54-232.callplus.net.nz)
12:00
12:04:31 Join kugel [0] (n=kugel@rockbox/developer/kugel)
12:10:33 Join {phoenix} [0] (n=dirk@p57AA6ACB.dip.t-dialin.net)
12:10:43 Join darko [0] (n=chatzill@c83-251-36-88.bredband.comhem.se)
12:11:10 Nick darko is now known as Guest25159 (n=chatzill@c83-251-36-88.bredband.comhem.se)
12:12:23 Join darko_ [0] (n=chatzill@c83-251-36-88.bredband.comhem.se)
12:12:32CIA-6New commit by 03Buschel (r24157): More tabs removed
12:13:42 Quit Guest25159 (Read error: 60 (Operation timed out))
12:14:45***Saving seen data "./dancer.seen"
12:15:05 Join darko21 [0] (i=53fb2458@gateway/web/freenode/x-cdlswpmgirexxbxy)
12:17:51darko21Does installing rockbox on my D2 using the rockbox utility give me multiboot? and does it format the flash memory?
12:18:05darko21<−− afraid
12:18:15darko21<−− backed up, but still
12:21:06 Quit darko_ (Read error: 60 (Operation timed out))
12:24:57 Join kaniini [0] (n=kaniini6@dyn75-70.yok.fi)
12:26:15 Join petur [0] (n=peter@d54C6F9B2.access.telenet.be)
12:27:36 Quit stooo (Read error: 60 (Operation timed out))
12:27:54 Quit kramer3d ("Leaving")
12:30:48 Join stooo [0] (n=sto@g226200059.adsl.alicedsl.de)
12:31:01 Quit bmbl (Read error: 104 (Connection reset by peer))
12:31:16CIA-6New commit by 03Buschel (r24158): Final removal of tabs in firmware path
12:48:22darko21Using "install rockbox" in the rockbox utility, the .rockbox directory is created on the mp3player drive, but it is read-only (the NAND driver issue) can I just move the .rockbox dir to the SD?
13:00
13:04:12pixelmaon a D2?
13:06:32darko21yeah
13:06:32 Join rhodan [0] (n=quassel@88-40.3-85.cust.bluewin.ch)
13:07:46darko21I apparently messed up while installing the bootloader, trying again. Didnt notice the "unknown OF file used" error :D
13:10:03 Quit S_a_i_n_t ("S_a_i_n_t [St.] has left")
13:10:08 Join S_a_i_n_t [0] (i=S_a_i_n_@203.184.0.57)
13:10:24 Quit S_a_i_n_t (Client Quit)
13:11:16darko21Nope, cant use the multiboot D2_2.59.d2 from the "HOW TO: Install Rockbox" thread on iaudiophile it seems
13:11:42 Join S_a_i_n_t [0] (i=S_a_i_n_@203.184.0.57)
13:12:16 Quit Utchybann ("I like core dumps")
13:15:38 Quit Buschel (Read error: 110 (Connection timed out))
13:23:13 Quit merbanan (Read error: 110 (Connection timed out))
13:27:58 Quit rhodan (Remote closed the connection)
13:30:27 Join rhodan [0] (n=quassel@88-40.3-85.cust.bluewin.ch)
13:36:22darko21I can't boot rockbox :( The bootloader install was successful and the rockbox utility shows rockbox (base) r24158 installed. Power then hold doesnt work, am I a noisy noob or is something wrong
13:36:22 Quit stooo (Read error: 104 (Connection reset by peer))
13:36:27 Join stooo [0] (n=sto@g226200059.adsl.alicedsl.de)
13:41:52darko21oooo crap
13:42:18darko21its not responding anymore
13:42:32 Quit kaniini ("BRB! IRL!")
13:43:35darko21not even to pluggin in the usb cable
13:45:17 Join einhirn [0] (i=Miranda@vpn10.rz.tu-clausthal.de)
13:50:05 Quit rhodan (Read error: 113 (No route to host))
13:52:11*amiconn found the reason for #10412 :)
13:52:37amiconnIt only happens on coldfire targets with sw tone control
13:52:57pixelmadarko21: turned the unit off before plugging the USB? Unfortunately I don't know much about the D2 and don't think there are people around currently how know more...
13:53:07amiconnThat's why irivers are normally unaffected. I was able to reproduce it with a special iriver build (sw tone controls enabled)
13:53:08pixelmas/how/who
13:54:02amiconnThe problem is that the bitmap scaler (used for album art) changes the emac mode, and somehow the tone filter relies on it being set properly (unlike the eq. which sets it)
13:54:30amiconnNow I need to find out how to handle this properly
13:57:23 Join Sajber^ [0] (n=Sajber@h-142-12.A213.priv.bahnhof.se)
14:00
14:01:21CIA-6New commit by 03bertrik (r24159): chip8 plugin: ...
14:02:08 Join Buschel [0] (n=ab@p54A3C388.dip.t-dialin.net)
14:03:03 Join kaniini [0] (n=kaniini6@dyn75-70.yok.fi)
14:07:32CIA-6New commit by 03mcuelenaere (r24160): Onda VX747: fix USB charging detection
14:14:48***Saving seen data "./dancer.seen"
14:14:52CIA-6New commit by 03bertrik (r24161): apps/gui/bitmap/list.c: use bool instead of int to hold boolean value
14:15:35bertrikI noticed the warning on the sim builds for the chip8 plugin, I'll fix it in a few minutes
14:16:06 Join fdinel [0] (n=Miranda@modemcable235.127-131-66.mc.videotron.ca)
14:16:45amiconnUnhelpful: Is the bitmap loader called from the audio thread?
14:18:05 Join Barahir_ [0] (n=jonathan@gssn-5f756f79.pool.mediaWays.net)
14:20:10 Quit Barahir (Read error: 60 (Operation timed out))
14:20:44CIA-6New commit by 03bertrik (r24162): Fix warning in chip8 plugin for simulator builds
14:32:18CIA-6New commit by 03mcuelenaere (r24163): Disable speaker when recording to prevent feedback.
14:34:35CIA-6New commit by 03amiconn (r24164): Coldfire targets: Restore EMAC status after using the EMAC for bitmap resizing. Fixes FS #10412. The tone and EQ filter calculation relies on the ...
14:36:03gevaertsbertrik: yellow
14:36:47 Quit Barahir_ (Connection timed out)
14:39:31gevaertsdarko21: you'll probably have to use tcctool for recovery now
14:46:47darko21gevaerts yup, I've fixed it with tcctool. Thanks :)
14:49:35 Join domonoky [0] (n=Domonoky@rockbox/developer/domonoky)
14:50:17darko21Oh for the love of jesus! I wasn't doing the last step of renaming the bin file in the root to D2N.bin, THATS why it didnt work. I. Am. Such. A. Noob.
14:51:01darko21sorry for the noise, I'll be quiet now
14:57:53 Quit einhirn ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
14:58:49 Join rhodan [0] (n=quassel@143-88.203-62.cust.bluewin.ch)
15:00
15:09:04 Quit {phoenix} (Remote closed the connection)
15:09:55 Join GeekShadow [0] (n=Antoine@reactos/tester/GeekShadow)
15:10:41 Join efyx [0] (n=efyx@lap34-1-82-225-185-146.fbx.proxad.net)
15:28:11CIA-6New commit by 03bertrik (r24165): Set svn:keywords property on .c and .h files that didn't already have it. Correct svn:executable property on some files.
15:34:19 Quit efyx (Remote closed the connection)
15:38:51 Join efyx [0] (n=efyx@lap34-1-82-225-185-146.fbx.proxad.net)
15:43:30 Quit efyx (Remote closed the connection)
15:44:53 Join efyx [0] (n=efyx@lap34-1-82-225-185-146.fbx.proxad.net)
15:48:00Unhelpfulamiconn: i'm late, aren't i? i believe it is..
15:52:45pixelmaUnhelpful: is the latter referring to amiconn's question or to your own? ;)
15:53:56Unhelpfuli stilled don't know the answer to my own question. :)
15:55:31Unhelpfuli suppose i could divide 0x7fffffff and 0x40000000 by 1 in a loop on beast to find out if unset bits in the result cost more. it doesn't really seem worthwhile since i know a much faster divider for that CPU...
16:00
16:09:35 Quit pjm0616 (":reboot")
16:09:49 Join JdGordon1 [0] (n=jonno@m465736d0.tmodns.net)
16:14:19 Join Sparky` [0] (i=Sparky@ip-83-134-87-223.dsl.scarlet.be)
16:14:50***Saving seen data "./dancer.seen"
16:18:16 Join CaptainKewl [0] (n=jason@207-237-117-89.c3-0.80w-ubr2.nyr-80w.ny.cable.rcn.com)
16:23:19 Join webguest36 [0] (n=3a6e8f25@giant.haxx.se)
16:23:32 Join StealthyXIIGer [0] (n=stealthy@69.216.116.162)
16:26:11 Quit webguest36 (Client Quit)
16:27:41 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
16:33:13 Quit JdGordon1 (Read error: 110 (Connection timed out))
16:36:38 Join perfectdrug [0] (n=5b0ee024@giant.haxx.se)
16:37:46 Quit petur (Remote closed the connection)
16:38:03perfectdrugpixelma: thanks for investigating and amiconn thanks for fixing FS #10412 this was fast after you could reproduce this bug, good work thanks
16:40:21 Join pjm0616 [0] (n=user@61.250.113.98)
16:43:37 Join WildFox500 [0] (n=chatzill@adsl-76-213-114-21.dsl.okcyok.sbcglobal.net)
16:49:39 Quit togetic (Read error: 110 (Connection timed out))
16:53:14 Join togetic [0] (n=togetic@unaffiliated/ibuffy)
16:54:39WildFox500Anybody home?
16:57:06CIA-6New commit by 03unhelpful (r24166): Invert divisor earlier in udiv32_arm, allowing the div0 test to be done before entering the 32-bit divide portion of the code, and making the handling ...
16:57:11CIA-6New commit by 03mcuelenaere (r24167): Onda VX747/VX777: disable LCD when backlight is off (saves power)
17:00
17:00:18 Join kugel_ [0] (n=kugel@e178083061.adsl.alicedsl.de)
17:00:32 Quit kugel (Nick collision from services.)
17:00:36 Nick kugel_ is now known as kugel (n=kugel@e178083061.adsl.alicedsl.de)
17:03:34 Quit StealthyXIIGer (Read error: 110 (Connection timed out))
17:06:40krazykitWildFox500, do you have a question?
17:06:56kugelUnhelpful: doesn't inverting the divisor mean that now all divisors that didn't have the top bit set now have it set?
17:08:37WildFox500I was wondering if I could use Rockbox to salvage my Gigabeat S. The factory firmware has tanked. It reads "Contact Toshiba for repair" as soon as it turns on.
17:09:58Unhelpfulkugel: actually, that comment needs a correction, it *does* work if the divisor has the top bit set, provided the numerator doesn't. also, it always inverted it, it just did it later on. the divisor is inverted so that you can subtract it from the remainder by adding. this also places the next bit of the result in carry, which is set if a result is non-negative...
17:18:07 Join Barahir [0] (n=jonathan@gssn-5f757c30.pool.mediaWays.net)
17:21:19 Join Lss [0] (n=Lss@cm23.omega219.maxonline.com.sg)
17:21:44 Join froggyman [0] (n=sopgenor@pool-72-69-220-194.chi01.dsl-w.verizon.net)
17:24:15 Join petur [0] (n=peter@d54C6F9B2.access.telenet.be)
17:25:06 Quit Buschel ()
17:28:18CIA-6New commit by 03mcuelenaere (r24168): Onda VX747/VX777: centralise LCD clock en-/disabling.
17:28:46 Part froggyman
17:34:58CIA-6New commit by 03mcuelenaere (r24169): Onda VX747/VX777: cleanup linker files a bit
17:35:57 Quit stoffel (Remote closed the connection)
17:41:22 Join stooo1 [0] (n=sto@g226200059.adsl.alicedsl.de)
17:41:23 Quit stooo (Read error: 54 (Connection reset by peer))
17:55:20CIA-6New commit by 03mcuelenaere (r24170): Onda VX747/VX777: update battery charge cycle
18:00
18:09:20 Quit darko21 ("Page closed")
18:13:27 Quit GeekShadow ("The cake is a lie !")
18:14:01 Join stripwax [0] (n=Miranda@87-194-34-169.bethere.co.uk)
18:14:53***Saving seen data "./dancer.seen"
18:23:32 Quit stooo1 (Read error: 104 (Connection reset by peer))
18:23:46 Join stooo [0] (n=sto@g226200059.adsl.alicedsl.de)
18:27:50 Join StealthyXIIGer [0] (n=stealthy@69.216.116.162)
18:29:01 Join toffe82 [0] (n=chatzill@adsl-71-132-87-108.dsl.sntc01.pacbell.net)
18:32:13CIA-6New commit by 03mcuelenaere (r24171): Onda VX747/VX777: fix not closing I²C all times
18:32:16CIA-6New commit by 03mcuelenaere (r24172): Onda VX747/VX777: only enable SD clock when transferring data
18:34:41 Quit WildFox500 ("ChatZilla 0.9.86 [Firefox 3.0.16/2009120208]")
18:47:24 Join MethoS- [0] (n=clemens@134.102.106.250)
18:52:29 Join stooo1 [0] (n=sto@g226200059.adsl.alicedsl.de)
18:52:29 Quit stooo (Read error: 54 (Connection reset by peer))
18:53:18 Quit Sparky` ()
19:00
19:03:25 Quit StealthyXIIGer (Read error: 110 (Connection timed out))
19:16:52 Join mirak [0] (n=mirak@ip-226.net-89-2-112.rev.numericable.fr)
19:21:47 Join Zarggg [0] (n=zarggg@65-78-69-194.c3-0.eas-ubr6.atw-eas.pa.cable.rcn.com)
19:25:24 Join merbanan [0] (n=banan@c-83-233-52-241.cust.bredband2.com)
19:25:50stripwaxsaratoga - this looks interesting. http://www.eurasip.org/Proceedings/Eusipco/Eusipco2006/papers/1568980508.pdf
19:29:01stripwaxBy rewriting the pre/post twiddle factors, you get more possibilities for merging the pre/post steps into the fft computation. It kindof sounds like it would only work well if we used a fixed radix rather than split radix fft though.
19:33:02 Join b1uebrother [0] (n=dom@f053153060.adsl.alicedsl.de)
19:33:44 Nick b1uebrother is now known as bluebrother (n=dom@f053153060.adsl.alicedsl.de)
19:35:03saratogastripwax: "The contribution of pre and post processing to the total computations in IMDCT varies from 34% to 79% depending its order."
19:35:08saratogadoes that sound wrong to you?
19:36:27stripwaxIncluding bitreverse, probably not very wrong, but for the sizes we use, it ought to be the very low end of that scale - no more than about 30% I would hope.
19:36:54saratogastripwax: IIRC when I benchmarked it the impact was small, just a couple MHz
19:37:04stripwaxThat's reassuring
19:37:31stripwaxBTW I'm thinking in terms of reducing the size of the tables/memory requirements rather than improving the pre/post mhz
19:37:43saratogaheh though maybe the rest of the algorithm perfomed badly enough that pre/post was negligable
19:37:55saratogayes I think that is more important
19:39:04saratogastripwax: what does the mdct_half trick actually save, a load and a store per sample or more? (ignoring memory savings)
19:39:56stripwaxi'm not super convinced it saves much at all unless you rewrite the pcm_buffer_insert to also do the windowing+reflection at the same time
19:40:49stripwaxotherwise you still need, at some point, the full N (not N/2), windowed, buffer somewhere.
19:41:18bluebrotherdomonoky: I'm thinking about providing rbutil binaries from svn every now and then. Do you think it's a good idea to announce that in the forums (unsupported builds?) as well?
19:41:37domonokybluebrother: sounds good.
19:42:59bluebrothershould the announcement in the unsupported builds forums, or in the rbutil forum as a sticky?
19:43:22bluebrotherI tend to vote for the rbutil forums.
19:44:23 Quit amiconn (Nick collision from services.)
19:44:24 Join amiconn_ [0] (i=quassel@rockbox/developer/amiconn)
19:44:33 Join einhirn [0] (i=Miranda@vpn10.rz.tu-clausthal.de)
19:44:44 Nick amiconn_ is now known as amiconn (i=quassel@rockbox/developer/amiconn)
19:45:32saratogastripwax: i think you save at least a load/store because you only have to write out N/2 samples at the end of the mdct, and only load N/2 in the windowing code
19:46:14stripwaxok true. is that really a massive saving?
19:46:23 Join pixelma_ [0] (i=quassel@rockbox/staff/pixelma)
19:46:23 Quit pixelma (Nick collision from services.)
19:46:40 Nick pixelma_ is now known as pixelma (i=quassel@rockbox/staff/pixelma)
19:46:47stripwaxit will make the windowing code much more complicated
19:47:25stripwaxYou'll need to load one sample, then multiply it by two different factors and write the results to two different places. With all N samples in the pcm buffer, the windowing code is very simple and fast.
19:48:24saratogastripwax: why two different factors?
19:48:41bluebrotherany opinions where to put such an announcement?
19:49:01saratogai thought the idea was that you are using the same factor on both and just accumulating them into different places
19:49:41stripwaxno. you are using different factors. the window is *not* symmetrical
19:50:04stripwaxfor vorbis at least, as far as I understand it.
19:50:55 Quit stooo1 (Read error: 104 (Connection reset by peer))
19:50:59 Join stooo [0] (n=sto@g226200059.adsl.alicedsl.de)
19:51:06 Quit rhodan (Remote closed the connection)
19:51:10saratogai didn't know that about vorbis
19:51:16saratogai thought windows were always symmetric
19:52:22 Part stooo
19:52:31Unhelpfuli thought they *had* to be for tdac? although i suppose the requirement might be different when overlapping blocks of differing sizes...
19:54:00saratogaUnhelpful: they don't have to be symmetric as long as they fullfill condition that the the two overlapped samples at each point have a square sum of unity
19:54:29saratogabut i think you're right that they typically are as long as the windows are the same size
19:54:30saratogahttp://www.xiph.org/vorbis/doc/Vorbis_I_spec.html#x1-260001.3.2
19:54:44saratogacertianly makes it look like the common case (long -> long) is symmetric
19:56:30stripwaxwait wait wait. we're talking about different things
19:56:37stripwaxThey are symmetric around the middle.
19:56:48stripwaxThe half-mdct is the region between N/4 and 3N/4
19:57:02stripwaxthe windows are not symmetric between [0,N/4] and [N/4,N/2].
19:57:17stripwax(neither are they symmetric between [N/2,3N/4] and [3N/4,N])
19:58:03stripwaxSo when you expand your half-mdct to a full mdct, you need to multiply by two different window factors, because your one load goes into two stores in different locations.
19:58:44stripwaxAnd as you say, no middle symmetry for long/short or short/long blocks
20:00
20:01:01 Quit fdinel ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
20:01:36saratogaso basically, the problem is that the window symmetry and the transform symmetry are about different points (N/2 and N/4,3N/4) and so you wouldn't be easily able to exploit that to save loads?
20:01:39stripwaxOr to put it another way: the right hand half of the 'full' mdct is not just a reflection of the left hand half of the 'full' mdct. To multiply by the same window factor and accumulate into two different places, it would have to be just a reflection I think.
20:01:55 Nick fxb__ is now known as fxb (n=felixbru@h1252615.stratoserver.net)
20:02:07stripwaxyes
20:02:18bluebrotherdomonoky: http://forums.rockbox.org/index.php?topic=23544.0
20:02:24saratogathat will make it a lot harder to exploit
20:02:29stripwaxI mean, you 'do' save loads, but at the expense of complexity in the writes
20:02:59stripwaxThe way it is now, you can just ldmia/stmia the whole lot
20:06:15 Quit mirak ("Ex-Chat")
20:06:16 Quit flydutch ("/* empty */")
20:07:11 Join fdinel [0] (n=Miranda@modemcable235.127-131-66.mc.videotron.ca)
20:09:06 Join funman [0] (n=fun@rockbox/developer/funman)
20:09:10 Join JdGordon| [0] (n=me@rockbox/developer/JdGordon)
20:14:57***Saving seen data "./dancer.seen"
20:15:00saratogastripwax: what if imdct_half multiplied the middle half by the correct window coefficients by folding them into tsin (or rahter making a new trig array thats tsin*window), and then the overlap code mirrored them and then multiplied just the mirrored coefficients by window/window_symetric
20:15:46stripwaxand divide by the already-multiplied coeffecients presumably?
20:15:54saratogayeah
20:16:26saratoganot 100% how well that would work with all the codecs though
20:16:46saratogaIIRC AAC is rather obnoxious about that
20:16:59saratogathough i suppose theres always mdct_full for the others
20:17:24stripwaxso then you definitely need a new trig table, and a separate trig table for long/short blocks, and you can't reuse the trig tables for fft, plus you need to change your codecs window tables to multiply by 1/the initial multiplications factors
20:17:47Unhelpfulwho knows gnu assembler well? i want to make an assembler macro that optionally omits or includes certain instructions if one of the macro arguments is unset or set to some sort of sentinel value... i'm working on a libgcc divider replacement and would like to macro-ize the main division body but have it store the remainder to r0 directly if it's being expanded into a mod function, or leave out instructions to finish the remainder if it's
20:17:47Unhelpful being expanded into a div function
20:17:49stripwaxor else do a real division, which is a nono
20:18:34Unhelpfuland for the eabi case the remainder and quotient are actually *both* needed
20:18:50saratogastripwax: does it take up more space, I was thinking that that the new tables would fit into the space occupided by the old window coefficients previously
20:19:35saratogayou only need N/8 of tsin and tcos right?
20:19:46Unhelpfulthe .if family seems promising but i'm a little lacking for examples :/
20:20:06saratogaso duplicate N/8 of each of those, and then duplicate N/8 of the window coefficients, but you then save all the space previously used for windowing
20:20:09kugelUnhelpful: both for eabi?
20:20:21saratogathough i'm probably not remembering how this actually works
20:21:27Unhelpfulkugel: eabi provides a divmod function that return the quotient and remainder in r0 and r1. this works great from a code-size perspective, really, especially with a divider that's updating both in one cycle anyway.
20:21:55stripwaxsaratoga - it sounds complex, and you won't be able to share the twiddle factors between mdct and fft. but then, it does sound like you might just be able to put the post twiddles and window factor into a single table, which is pretty neat
20:22:23stripwaxas in, I'm thinking it's not even about mdct-half, but you could just do that trick 'anyway', and save a multiply for every sample ...
20:23:23Unhelpfulthe .if variants only seem to process quoted strings or numeric expressions, though... so i'm not really sure what to do with a register name as an argument.
20:24:06saratogastripwax: looking at wma, we allocate a full N samples space for the window coefficents, so for that codec at least, you could store all the duplicated tcos/tsin and both new sets of window coefficients in the same space
20:24:35saratogathe way i see it you would still share tcos/tsin between mdct.c and fft-ffmpeg, its just you would only use them for pre rotation and not post
20:24:53saratogaFWIW wma can't even fit the window coefficients into IRAM
20:24:57stripwaxsaratoga - yep, me too (according to that pdf)
20:25:05stripwaxotherwise the prerotation factors are still different from the fft ones
20:25:29saratogayeah i think this is basically what they wanted to do, though i admit i didn't understand half the paper
20:25:35 Quit Tomis (Read error: 110 (Connection timed out))
20:27:17Unhelpfulah, wait, modern binutils docs read differently: http://sourceware.org/binutils/docs/as/If.html#If
20:28:29stripwaxActually i'm unusre- I think if you did multiply the middle-half in the imdct first, you wouldn't be able to the reflection later (unless you divide). sounds bad. That probably also means you can't combine the post twiddle and window factors into a single table, since the reflection happens *between* post twiddle and windowing
20:30:19 Quit JdGordon| ()
20:50:18 Join steve79 [0] (n=Steve@h-74-0-49-42.cmbrmaor.static.covad.net)
20:52:57 Join stripwax_ [0] (n=Miranda@87-194-34-169.bethere.co.uk)
20:54:41steve79I'd really like to try rock box and am trying to figure out what supported players can be readily purchased and which are the of those are the best. I've been doing some investigation but figured I'd find some valuable insight here.
20:55:14 Join Horscht [0] (n=Horscht2@xbmc/user/horscht)
20:56:16 Quit fdinel (Read error: 104 (Connection reset by peer))
20:57:36steve79I like the fuze but it seems like it will be impossible to figure out whether used or refurb players are v1 prior to purchasing.. started looking into the e280 but wasn't sure about support. Some places mention e200 series but the hardware page talks only about the e200.
20:59:32gevaertse200 always refers to the entire series
21:00
21:01:34steve79Got it, thanks. Do regrets await me with the e280 or does that seem like a pretty good choice?
21:03:31gevaertsit all depends on what you're looking for
21:04:27*gevaerts hasn't actually used an e200 seriously, so he isn't the best one to offer advice
21:04:47CIA-6New commit by 03tomers (r24173): rbutil: Explicitly set some widgets layout direction to LTR
21:05:38steve79Basically something I can get rockbox on. Large storage, scroll wheel (especially one that actually moves), and thin are serious pluses.
21:06:02funmane200v2 & fuze lcd controller look identical
21:06:28 Join tomers [0] (n=chatzill@bzq-84-109-85-100.red.bezeqint.net)
21:07:17gevaertssteve79: I think you'd probably be happier with a fuze, but as you said, it's a bit tricky to get a v1
21:07:20funmanthe main difference is the screen orientation is modified in mpegplayer
21:07:31funmanon e200v2, and not on fuzer
21:07:32funman-r
21:08:38gevaertssteve79: the e200 isn't bad though
21:08:38 Quit petur (Read error: 54 (Connection reset by peer))
21:09:03 Join fdinel [0] (n=Miranda@modemcable235.127-131-66.mc.videotron.ca)
21:09:23 Join petur [0] (n=peter@d54C6F9B2.access.telenet.be)
21:10:35 Quit stripwax (Read error: 110 (Connection timed out))
21:13:27steve79gevaerts: is this a trick to getting a v1 or just luck?
21:14:44gevaertssteve79: if you can get the seller to check the firmware version, do so. If you can't, it's luck
21:15:23gevaertsmarkings on the case are meaningless
21:16:59 Quit Grahack ("Tu m'as vu ?")
21:17:39CIA-6New commit by 03tomers (r24174): rbutil: Fix some messages
21:18:52funmanrockbox works quite well on e200v2 (just like on fuzev1)
21:21:19 Quit stripwax_ ("http://miranda-im.org")
21:24:28bluebrothertomers: why is removing a full stop at the end of a message "fixing" it?
21:25:05bluebrotherthis is a sentence, to IMO the full stop does make sense and is correct.
21:28:12bluebrotherand putting a space in front of an ellipsis isn't wrong in english either.
21:28:20 Join einhirn_ [0] (i=Miranda@vpn10.rz.tu-clausthal.de)
21:28:36bluebrotherhaven't checked what's more common, but at least it's correct.
21:28:45bluebrothertomers: I really disagree with this "fix"
21:29:19steve79I think I'm ready to get a player.. thanks a lot for your help gevaerts and funman!
21:36:23 Join StealthyXIIGer [0] (n=stealthy@69.216.116.162)
21:41:34 Join stooo1 [0] (n=sto@g226200059.adsl.alicedsl.de)
21:41:44 Join BHSPitLappy [0] (n=BHSPitLa@unaffiliated/bhspitmonkey)
21:41:48 Quit liar (Read error: 104 (Connection reset by peer))
21:42:03 Join liar [0] (n=liar@83.175.83.185)
21:45:52 Quit petur (Remote closed the connection)
21:49:39 Join stooo [0] (n=sto@g226200059.adsl.alicedsl.de)
21:49:39 Quit stooo1 (Read error: 54 (Connection reset by peer))
21:49:57 Quit xavieran (Connection timed out)
21:50:19tomersbluebrother: The full stop is not consistent. The messages dialog box contained some strings with full stop, some without. Doesn't it acceptable that list items does not have full stop? If you like, I can use full stop on all the items (all or none approach)
21:51:12bluebrotherI agree that making it consistent is a good thing but in that case we need to agree on which route to go _first_.
21:51:22bluebrotherplus, a space in front of an ellipsis is correct.
21:52:08tomersbluebrother: Isn't it accepted that e.g. bulleted lists (like in presentations) does not contain stops?
21:52:09bluebrotherideally we should figure what's the common way.
21:52:56bluebrotherwell, the question here is what's commonly used. Of course one can omit stops after sentences, but at least for writing this is simply wrong.
21:53:08tomersbluebrother: Although English is not my primarily language, I thought this well known. If I am mistaken, then I apologize
21:53:22bluebrotherso instead of simply removing stops we should first figure what's the common way of doing it.
21:54:08tomersThese string are mostly (or only) used in these dialog boxes. So let's decide on the way, and I will fix this
21:54:31bluebrotherbesides, calling such a change "fix" is somewhat lying. It doesn't fix anything, the strings were correct before. It's making them consistent.
21:54:47tomersagree
21:55:09bluebrotherwell, ideally someone speaking english natively could tell us what's the common way of doing this for programs.
21:55:36 Join kugel_ [0] (n=kugel@e178083061.adsl.alicedsl.de)
21:56:02tomersbluebrother: I might go soon, so I'll read the IRC log tomorrow, and fix it tomorrow. Are you going to discuss this with the others, or should I do so tomorrow?
21:56:11 Quit einhirn (Read error: 110 (Connection timed out))
21:57:45bluebrotherwell, it seems to be rather quiet right now. I guess most are still on vacation or similar. We'll see who's faster ...
21:57:59tomersok... good night
22:00
22:00:03funmanthere are some registers used in fuze lcd which aren't used on e200v2 and the reverse is also true
22:00:40funman2 are likely power registers, and 5 are specific to fuze windowing
22:01:30kugel_funman: I found you can actually change the contrast on the fuze with some registers that the e200v2 sets on initializing, but I haven't found a way to make it useful
22:01:50funmaninvert & flip works on fuze
22:02:07funmani'll post an update with this on 10603
22:02:55bertrikif you add invert for fuze, I'll add it for the c200/c200v2 ! :P
22:03:43kugel_does anybody have an idea why fs10603 doesn't work on the e200v2?
22:04:06funmani was trying to merge fuze & e200v2 lcd drivers to see if the difference was there
22:04:55kugel_I thought that the e200v2 can still not read the wheel during lcd updates, but then again it works for the fuze's home&power button which also cannot be read reliably during updates (with the svn code)
22:05:49kugel_and hold, that one gives false positives without the precharging for example
22:06:10 Quit tomers ("ChatZilla 0.9.86 [Firefox 3.5.6/20091215231400]")
22:10:24 Quit pamaury ("exit(*(int *)0 / 0);")
22:11:36 Quit kugel (Read error: 113 (No route to host))
22:12:20funmanbertrik: do you still have that mkamsboot c200v2 diff ?
22:12:54bertrikfunman, that doesn't ring a bell, so probably not
22:14:59***Saving seen data "./dancer.seen"
22:22:26 Quit StealthyXIIGer (Read error: 60 (Operation timed out))
22:25:18funmanbertrik: can you test http://pastie.org/765275 ? it reads c200v2 buttons with dbop instead of gpio
22:28:55 Quit Guest16836 (Client Quit)
22:29:07 Quit stooo (Read error: 104 (Connection reset by peer))
22:29:08 Join AlexP [0] (n=alex@rockbox/staff/AlexP)
22:29:31 Join stooo [0] (n=sto@g226200059.adsl.alicedsl.de)
22:31:43kugel_funman: maybe it would be better to put the unification into a different patch and commit it beforehand
22:32:35funmantrue
22:33:34evilnickIs anyone around with an e200 here?
22:34:50funmanmc2739: can you test fs#10603 with matrix demo ? it's lcd intensive and uses the wheel
22:36:20evilnickIt's probably repeatable on other targets too, but if the user plugs USB in after starting the very first database scan then the scan will pause on wherever it got up to but then not update any further when USB is unplugged again
22:37:29kugel_that doesn't surprise me
22:38:41kugel_funman: I was actually planning to look at unification once the dbop patch is in
22:39:09funmankugel_: i notice you changed lcd_window_{x,y} to lcd_window() on e200v2 in r20063
22:39:33funmando you remember why ?
22:40:29funmanlcd_window_x() / lcd_window_y() calls aren't sequential in fuze' lcd_blit_yuv so there is a gain of having them splitted
22:41:18funmanthough e200v2 has a special lcd_window_blit() for lcd_blit_yuv()
22:41:21 Quit einhirn_ ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
22:41:30kugel_the split version wasn't correct
22:41:40kugel_note the last line of lcd_window()
22:42:31gevaertsevilnick: that's probably going to happen on all targets, including hardware usb
22:43:39funmani'll look at committing the diffs to fuze and e200v2 (only cosmetics since I can't test it)
22:43:44funman+ tomorrow
22:44:17evilnickgevaerts: Wouldn't it be nicer (for novice users) to have a system whereby there's a flag set for database initialisation is running that gets cleared once it's completed, and that way if the process is interrupted then it can be re-run after the interruption?
22:44:21 Quit funman ("free(random());")
22:44:45 Quit merbanan (Read error: 110 (Connection timed out))
22:45:14gevaertsevilnick: probab.y, yes
22:45:44kugel_evilnick: we haven't had a single complaint about that, so I'm not sure if it would be really better
22:46:44gevaertsevilnick: did you start the scan by hand?
22:46:55evilnickkugel_: Yes, it'd be a very small fraction of the userbase, but that's the kind of attention to detail that I've come to expect from RB :)
22:47:39kugel_it could potentially annoy a bigger fraction of userbase
22:48:01evilnickgevaerts: It was on the very first run on a new build so I went to Database > start scanning (or w/e the term is)
22:49:12evilnick"Initialize now"
22:51:08*gevaerts isn't sure
23:00
23:03:40 Join stempniak [0] (n=63f3ecf9@giant.haxx.se)
23:04:17stempniakhello, how do I make my player go in to charging mode, instead of USB keypad mode
23:08:14 Quit perfectdrug ("CGI:IRC")
23:09:00stempniakHello, when I plug Sans e250 in usb, how do I prevent it from going to usb keypad mode instead of usb charging mode. Thanks
23:09:46gevaertsstempniak: hold the appropriate button while pluging in. I thing it's select on sansa, check the manual to make sure
23:10:36stempniakI thought you don't need to enter OF to charge in rockbox
23:10:43stempniakIm pluggin it in while rockbox is on
23:10:57gevaertsyes
23:12:32stempniakoh cool select works
23:14:03stempniakthanks
23:15:17 Quit stempniak ("CGI:IRC")
23:17:22 Part stooo
23:17:41 Quit bertrik ("De groeten")
23:18:44 Join Strife89 [0] (n=michael@adsl-146-208-54.mcn.bellsouth.net)
23:21:56 Join StealthyXIIGer [0] (n=stealthy@69.216.116.162)
23:31:25 Quit dfkt ("-= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.")
23:32:51 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
23:35:40 Quit StealthyXIIGer (Read error: 60 (Operation timed out))
23:49:07 Join perfectdrug [0] (n=5b0ee024@giant.haxx.se)
23:52:43 Quit perfectdrug (Client Quit)
23:53:02 Quit bmbl ("Bye!")

Previous day | Next day