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 2009-12-13

00:01:20 Quit liar (Read error: 104 (Connection reset by peer))
00:04:13***Saving seen data "./dancer.seen"
00:07:36 Quit mister_pink ("Leaving")
00:10:10 Join Sajber^1 [0] (
00:12:21 Quit yawny (Remote closed the connection)
00:12:27 Join elcan [0] (
00:14:30 Quit pamaury ("exit(*(int *)0 / 0);")
00:16:54CIA-6New commit by Domonoky (r23958): try to improve the new column add function.
00:18:24CIA-6New commit by Domonoky (r23959): commit also db class modifications.
00:23:14CIA-6New commit by Domonoky (r23960): revert broken checkwps.
00:25:51 Quit Sajber^ (Read error: 110 (Connection timed out))
00:31:34 Join liar [0] (n=liar@
00:35:48CIA-6New commit by Domonoky (r23961): Make sure to keep all rowids the same when adding a column.
00:36:28 Quit liar (Client Quit)
00:39:37 Quit BHSPitLappy (Remote closed the connection)
00:39:37 Quit CaptainKwel (Read error: 104 (Connection reset by peer))
00:43:26archivator_Can anyone think of a good reason why when I use greylib, I end up with an entire screen of black, although I never draw to the whole screen (except for a clear_display() but that's not really drawing..) ?
00:47:36CIA-6New commit by Domonoky (r23962): commit the download count patch again.
00:48:27 Join fyrestorm [0] (
00:50:35 Quit kugel (Read error: 110 (Connection timed out))
00:51:37domonoky1archivator_: maybe greylib is not correctly inited ?
00:51:55 Part tg1
00:53:11archivator_domonoky1: it is. Apparently LCD_DEFAULT_FG and LCD_DEFAULT_BG are 0 3 respectively :-/ Had to do GREY_BLACK, GREY_WHITE explicitly..
00:54:20domonoky1archivator_: yes, LCD_DEFAULT_FG/BG are values for the native grey level (on grey targets)
00:54:48archivator_Well, how was I supposed to figure that out? :O
00:55:02domonoky1reading code :-)
00:55:16archivator_I prefer docs :(
00:56:06domonoky1feel free to help document it in the wiki.
00:57:21 Join togetic [0] (n=togetic@unaffiliated/ibuffy)
00:58:24archivator_Will probably do, once I'm done with my plugin :)
00:58:27 Quit Omlet ("( :: NoNameScript 4.22 :: )")
00:59:06pixelma ?
01:00:14archivator_pixelma: not really greylib specific.. doesn't say anything about initialization, for example..
01:00:24pixelmadoes someone have a c200 at hand?
01:00:40pixelmawith a fairly recent build
01:02:23 Quit TopyMobile (Read error: 104 (Connection reset by peer))
01:04:00 Quit Highlander ("Quitte")
01:06:56*domonoky1 made download counts on theme page live :-)
01:10:26 Join mcuelenaere [0] (n=mcuelena@rockbox/developer/mcuelenaere)
01:11:52saratogahow come the onda doesn't have a picture on the themes site?
01:12:05mcuelenaeresaratoga: simple, because no-one has made one
01:13:37saratogaalso why has themes that obviously will not work on the clip
01:13:46*archivator_ got his plugin to work with greylib! \o/
01:15:14archivator_domonoky1: you could change the label to "Downloaded X time(s)" instead of "Download count" :)
01:15:48mcuelenaeredomonoky1: haven't looked at the implementation, but why do you need to pass an target= parameter in the index.php?themeid= URL? Can't you get that from the DB?
01:15:59archivator_And thanks for getting the content-type right :)
01:16:03mcuelenaerealso, it looks like you can edit that url to fit any theme to any target
01:16:24*domonoky1 goes changing the texts.
01:16:53saratogahow does the theme site figure out which themes go where?
01:17:05*mcuelenaere thinks he knows the reasoning for the target= param
01:17:27domonoky1mcuelenaere: the target parameter is just used to construct the correct link back. but i cant get this from the db, because a theme might fit multiple targets.
01:17:28mcuelenaerestill, allowing any value is a bug (although not a big one)
01:17:39mcuelenaeredomonoky1: yes, I just figured that out :)
01:18:13domonoky1it allows any value, and even shows any theme (even hidden) if you know the id, but i dont consider that as a bug. it doesnt harm.
01:20:05mcuelenaeredisplaying hidden themes is pretty bad, isn't it? Wasn't (part of) the point of hidden themes to "remove" themes with license problems from the site?
01:20:34mcuelenaereNow (people who use) search engines will still be able to acces it
01:21:24pixelmasaratoga: about the clip themes - my guess is that the ones that look misplaced are Iriver themes with RWPS (the Iriver lcd remotes have the same screen resolution as the Clip)
01:22:40 Quit mcuelenaere ("Off to bed..")
01:22:43 Join martian67 [0] (n=arkfar@about/linux/regular/martian67)
01:22:58pixelmastill not correct though
01:23:18pixelmathis reminds me to post my wishes to the forum thread.... but I don't want to scare people off ;)
01:30:04 Join anewuser [0] (i=anewuser@unaffiliated/anewuser)
01:32:43pixelmathe three that show up there do have .rwps files - and on the M3 page are only X5 themes with Iaudio rwps (the only reason there are themes for it at all)...
01:40:21saratogaanyone know what the S3 status is?
01:46:44 Quit robin0800_ (Read error: 60 (Operation timed out))
02:00:50CIA-6New commit by Domonoky (r23963): reword the download counter, and dont show themedetails for hidden or unapproved themes.
02:04:14***Saving seen data "./dancer.seen"
02:08:32CIA-6New commit by Domonoky (r23964): ups, fix single theme display.
02:12:39 Join chrisb [0] (
02:14:21chrisbi read the FAQ. it seems that rockbox devices work with docking stations on a case-by-case basis
02:15:02chrisbwith a mixture of recharging, audio out, etc.
02:18:00anewuserhmm, I plan to do an automated install of the rockbox firmware
02:18:26anewuserthing is my sansa c250 model didn't come with teh usb setting out of the box
02:18:49anewuserso I'm thinking the automated install will fail due to not having the MSC mode
02:18:58anewuserwould that be a proper assumption?
02:19:31krazykitanewuser, i believe this is covered in the wiki, actually
02:19:51anewuseroh, missed it
02:19:56anewuser*goes to check*
02:21:32 Join JdGordon [0] (n=jonno@rockbox/developer/JdGordon)
02:34:41 Quit Rondom (Nick collision from services.)
02:34:44 Join Rondom [0] (
02:47:22 Quit Lss (Read error: 104 (Connection reset by peer))
02:48:14Unhelpfuln1s: it appears that we can't actually merge ffs and log2 tables, they're not sparse *enough*. it would still be a much smaller table (64B) than the one used by bs_generic, though.
02:52:41saratogawhat is ff?
02:55:17Unhelpfulsaratoga: existing L_ffs_table used in core by find_first_set_bit. a very similar algorthim can be used to calculated floor(log2(x)) at a decent cost/space tradeoff between a straight binary search without LUT and one that finishes via a large LUT.
02:55:36saratogaUnhelpful: which codec?
02:57:43Unhelpfulwe were thinking it might be reasonable to make an exception to the one-log2-to-rule-them-all code for that case, as that log2 uses a rather large LUT of 256B
02:59:32CIA-6New commit by Domonoky (r23965): try to improve checkwps to not overwrite results if there is already a bad one.
03:01:09Unhelpfulsee ffs-arm.S, or ffs-coldfire.S. the same algorithm could be applied to log2, but with different setup, multiplier, and table. x |= x >> 16; x |= x >>8; x |= x >> 4; x |= x >>2; x |= x >>1; x *= 0x572fa8d; return L_ilog2_tbl[x];
03:03:35 Quit Thundercloud (Remote closed the connection)
03:15:10 Quit anewuser (Client Quit)
03:16:46 Join webguest55 [0] (
03:17:22 Quit webguest55 (Client Quit)
03:35:32 Quit GeekShadow ("The cake is a lie !")
03:38:44CIA-6New commit by Domonoky (r23966): Another attempt at improving checkwps. Only check targets with matching remote/mainlcd depending on filetype. And remove all old checkwps results of a ...
03:44:29 Quit archivator (Read error: 110 (Connection timed out))
03:44:42 Quit archivator_ (Read error: 110 (Connection timed out))
03:44:50 Quit domonoky1 (Read error: 104 (Connection reset by peer))
03:55:47 Quit GodEater (Remote closed the connection)
03:55:59 Join GodEater [0] (
04:04:17***Saving seen data "./dancer.seen"
04:06:51 Join Tomis2 [0] (n=Tomis@
04:14:41 Quit Tomis (Read error: 110 (Connection timed out))
04:14:41 Nick Tomis2 is now known as Tomis (n=Tomis@
04:25:00 Join BHSPitLappy [0] (n=BHSPitLa@unaffiliated/bhspitmonkey)
04:41:12Unhelpfulamiconn: it looks like coldfire has similar behavior for register shift to ARM, although the manual says it's mod 64 instead of mod 256... so a shift by a negative number of sufficiently small magnitude should produce a 0?
04:43:57 Quit efyx_ (Remote closed the connection)
04:45:42CIA-6New commit by unhelpful (r23967): Libfaad pow2/log2 improvements: ...
04:48:47saratogaUnhelpful: is there any improvement in SBR performance with those changes?
04:57:02Unhelpfulsaratoga: very small on arm, on the order of about .5% on e200. coldfire will probably benefit a bit more as MUL_R is a rather larger chunk of asm. i *think* that the DESCALE_SHIFT used on ARM would work well on coldfire, but i haven't tested. it relies on shift-by-negative producing a 0 result
04:57:42saratogathen it sounds like SBR was calling pow too much, which is a good sign
04:58:14Unhelpfulwasn't, you mean?
04:58:24saratogasorry, "wasn't"
04:59:25Unhelpfuli also removed an explicit 0-input test from pow2... this occurs on the order of <.01% of the time in my test files, so paying two instructions for it on every call is probably wasteful
05:02:02Unhelpfulit might be worth labeling these functions as pure, also, and possibly moving them so that they can be inlined
05:03:54Unhelpfulthere are very few callsites, and all are in sbr_hfadj.c
05:06:36saratogaif they're used so infrequently its probably not worth it
05:06:48CIA-6New commit by unhelpful (r23968): Fix red, DESCALE_SHIFT macro was missing a right-parenthesis.
05:08:01Unhelpfulit might not pay off much, but "move some functions from common.c to sbr_hfadj.c and label them static inline" is a *very* small effort. :)
05:14:05 Part toffe82_
05:43:28 Join Tomis2 [0] (n=Tomis@
05:51:02 Quit Tomis (Read error: 110 (Connection timed out))
05:51:02 Nick Tomis2 is now known as Tomis (n=Tomis@
06:04:18***Saving seen data "./dancer.seen"
06:12:17 Quit fdinel ("Miranda IM! Smaller, Faster, Easier.")
06:12:50 Quit chrisb (Read error: 104 (Connection reset by peer))
06:40:57 Join krazykit` [0] (
06:45:52 Join CaptainKewl [0] (
06:56:32 Quit krazykit (Read error: 110 (Connection timed out))
06:58:09 Join Llorean [0] (
06:59:09 Quit Llorean (Client Quit)
07:00:17 Quit panni_ ("( :: NoNameScript 3.81 :: )")
07:12:38 Join bluebrother [0] (
07:13:59 Quit JdGordon ("Leaving.")
07:15:33 Join Horschti [0] (n=Horscht2@xbmc/user/horscht)
07:16:17 Join JdGordon [0] (n=jonno@rockbox/developer/JdGordon)
07:25:06 Quit bluebroth3r (Read error: 110 (Connection timed out))
07:32:26 Quit DogBoy ("Leaving")
07:33:39 Join Llorean [0] (
07:33:58 Quit Horscht (Read error: 110 (Connection timed out))
07:53:31 Quit CaptainKewl (Remote closed the connection)
07:56:07 Quit Llorean ("Leaving.")
08:00:55 Quit BHSPitLappy (Read error: 110 (Connection timed out))
08:04:21***Saving seen data "./dancer.seen"
08:10:42 Join Grahack [0] (
08:29:15 Join n1s [0] (n=n1s@rockbox/developer/n1s)
08:31:09 Quit saratoga ("Page closed")
08:49:34n1sUnhelpful: "i *think* that the DESCALE_SHIFT used on ARM would work well on coldfire, but i haven't tested. it relies on shift-by-negative producing a 0 result" should i test that? have my h300 right here
09:00:46 Quit phanboy4 (Read error: 110 (Connection timed out))
09:04:16amiconnn1s: There are definitely no fonts with more than 64K glyphs, for two reasons. (1) Rockbox only supports the unicode BMP. (2) Same applies to our source font format, BDF
09:04:33 Join Highlander [0] (
09:05:28n1samiconn: aha, but still, the buffer these guys are increasing is limited to 60kB, right?
09:05:46n1sas in bytes, not glyphs
09:05:57n1sor am i way off here?
09:06:12amiconnEh, what guys?
09:06:50n1samiconn: FS #10844
09:08:05n1sthere was some discussion of it yesterday, there seems to be an issue where increasing the buffer fixes major slowness on boot when using what was called "non-latin" fonts
09:08:43n1sbut increasing the buffer to more than 64kB breaks our font code
09:10:42amiconnA font that doesn't fit completely will always be slower than a font that does at boot, because of the lru caching
09:10:44n1sthere's a patch in fs#7045 but it looks like a hack and i don't quite understand it to be honest as it just casts the pointer that indexed addressing is done on from short* to long* so i would think it would read from the wrong place but i am not too familiar with this
09:11:28n1samiconn: but should it take like 30 seconds to do that caching? and isn't the glyphcache supposed t be saved?
09:11:32amiconnUmm, the link to that patch is struck out?
09:11:42n1sit's been closed
09:11:58amiconnDunno, perhaps some things happen in the wrong order
09:13:39n1swhat do tou think of making the font buffer size configurable like the max files in dir?
09:13:57n1s(with said size limiting bug fixed)
09:15:50amiconnI don't really like those manually configurable limits. In some cases they are still necessary, unfortunately
09:16:20n1snor do i
09:17:07amiconnMany users don't really think about what to set there. They either don't know/care at all, or set minimal or maximal values at will leading to poor performance and/or strange effects
09:17:21n1si'd still like it fixed so people who make custom builds can change it easily
09:19:30 Join Grahack1 [0] (
09:19:35 Quit Grahack (Read error: 110 (Connection timed out))
09:20:36amiconnThe mentioned fix fixes an actual bug in the font code and looks like it's correct
09:20:45amiconnImo it should be applied
09:21:14n1sah, can you explain it, as i don't get it at all :)
09:21:33n1salso, you have commit access, dont' you?
09:21:52amiconnWhen markun added unicode support, our font format needed to be extended in order to support fonts > 64KB
09:22:07amiconnThe original format used 16 bit offsets
09:22:53amiconnIn order to keep old fonts compatible, he decided to support both 16 bit and 32 bit offsets, based on the total font size
09:23:12n1saha, that explains it :)
09:23:33amiconnThis check for pf->bits_size < 0xFFDB already happens in several places in font.c
09:27:10n1soh, and i guess the cast should be to uint32_t* instead of long* or 64 bit sims will be broken again
09:27:12amiconnAh yes. The bug hits if a fully loaded font is accessed which is larger than 64KB
09:27:51amiconnfont_get_bits() simply assumes this font is smaller than 64KB atm
09:28:23amiconnIf we want to save a bit of binsize, the fix could be ifdefed on MAX_FONT_SIZE
09:29:26amiconnFor consistency we should probably use unit16_t instead of unsigned short as well
09:29:43amiconnuint16_t too
09:30:19 Join petur [0] (
09:32:00amiconnThe logic could be reordered a bit so that the fixed-width font case isn't duplicated
09:33:17 Join antil33t [0] (
09:41:01*amiconn thinks that pf->offsets should probably be a void* in order to make its dual mode use more obvious
09:41:25n1si agree with that, also a comment would be nice :)
09:42:16amiconnThe 0xFFDB magic number should also be replaced by a descriptive macro
09:44:08 Join flydutch [0] (
10:00:00n1samiconn: are you working on this or do i have a project for this morning?
10:03:10 Join stoffel [0] (
10:04:25***Saving seen data "./dancer.seen"
10:07:29amiconnI have no time to work on this right now
10:13:48n1si'll look at it then, thanks for the hints
10:22:25 Join nick-p [0] (
10:22:46nick-pHi, I've written a patch (FS #10849) that supersedes one I wrote earlier (FS #10779). Do you know how I'd go about closing the old one?
10:26:59 Join pamaury [0] (
10:31:19 Join Kitar|st [0] (
10:37:43n1snick-p: ask in here :)
10:40:06nick-pCan FS #10779 be closed please :)
10:40:21n1ssure :)
10:43:41n1snit the future though, i would recomend posting a new version in the same task
10:44:43nick-pOK. I should have asked here first. Maybe a rename of the title would have been enough.
10:46:41Unhelpfuln1s: if you want to. just replacing the DESCALE_SHIFT macro with the inline version used on ARM *should* work without changing the output, which you could fairly easily test via test_codec. it might need some tweaking to be faster, or the existing one might be fairly optimal.
10:47:08n1sUnhelpful: ok, ill do that then
10:48:17Unhelpfuli also think some shuffling about of registers could allow using the same trick in ffs-coldfire.S that is used in bs_generic to load the result over the index and avoid having to zero-extend the result. as in bs_generic the high 24 bits of the index are already clear.
10:48:28 Quit Kitr88 (Connection timed out)
10:48:45 Join Kitr88 [0] (i=Kitarist@
10:49:25 Quit Kitar|st (Connection timed out)
10:56:58n1sUnhelpful: that is a pretty nice trick :)
11:00:41 Quit fyrestorm (Read error: 131 (Connection reset by peer))
11:02:10 Join fyrestorm [0] (
11:10:05 Join liar [0] (n=liar@
11:12:16n1shmm MAX_FONTSIZE_FOR_16_BIT_OFFSETS is a little long...
11:20:20 Quit nick-p ("CGI:IRC (EOF)")
11:27:38 Join webguest64 [0] (
11:28:46 Quit Grahack1 (Read error: 145 (Connection timed out))
11:28:59webguest64am I alone?
11:30:40 Join Grahack [0] (
11:32:24 Quit webguest64 (Client Quit)
11:32:34 Join webguest40 [0] (
11:34:08 Quit webguest40 (Client Quit)
11:43:27 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
11:54:05 Join Tomis2 [0] (n=Tomis@
12:04:26 Quit Tomis (Read error: 110 (Connection timed out))
12:04:26 Nick Tomis2 is now known as Tomis (n=Tomis@
12:04:29***Saving seen data "./dancer.seen"
12:07:43CIA-6New commit by nls (r23969): Font improvements: Fix bug that caused some fonts to be rendered garbled on custom builds with MAX_FONT_SIZE > 64k (closes FS #10844). Simplify version ...
12:07:44 Nick YPSY is now known as Ypsy (
12:07:59bluebrotherhmm. Theme installation is broken in rbutil svn due to the changed target names :/
12:10:22 Join Omlet [0] (
12:15:38bluebrothern1s: already spotted the red?
12:16:19n1shave now
12:17:25 Join kood777 [0] (
12:19:08CIA-6New commit by nls (r23970): fix typo causing red
12:37:42 Quit Grahack (Read error: 60 (Operation timed out))
12:38:04 Join Grahack [0] (
12:43:23 Join tg1 [0] (n=thomas@
12:52:15 Join MethoS- [0] (n=clemens@
12:58:48 Join teru [0] (
13:06:40 Quit kood777 ("CGI:IRC (EOF)")
13:17:43 Quit Grahack (Read error: 110 (Connection timed out))
13:20:09 Join Grahack [0] (
13:34:36 Join domonoky [0] (n=Domonoky@rockbox/developer/domonoky)
13:35:37 Quit liar (Read error: 104 (Connection reset by peer))
13:36:10 Join archivator [0] (n=archivat@
13:42:34 Join stooo [0] (
13:43:03CIA-6New commit by rasher (r23971): Spelling and grammar fix
13:47:46 Join Lear [0] (i=chatzill@rockbox/developer/lear)
13:50:15domonokypixelma: i just read your themesite whishes. The problem with the rwps as main wps is, that currently that wont work because the theme config set it as remote wps ofcourse. (so it will only work if the uses explicitly chooses to set the rwps file as wps manually).
13:52:27domonokyThats why i changed the rwps handling for now, to not show themes where a rwps fits a mainscreen.. of course more independend remote theme handling would be nice, but that gets complicated. I would prefer if uploaders just repackage the rwps into a seperate themezip if it can be also used as a mainwps.
13:53:08pixelmaI know that the cfg is a bit problematic there - and as I said, maybe it *could* work similar to packaging themes in SVN where the cfg file is created when making the zip. I guess using RWPSs as main WPSs could work more easily than the other way around
13:53:48 Join archivator_ [0] (n=archivat@
13:55:11pixelmaif the authors could get a list of targets, your suggestion would be a start
13:55:45domonokyhm, i dont really want to repack and rewrite the config on the theme site (currently we dont even check the contents of the config, because its difficult to define whats legal).
13:56:56pixelmaand maybe if someone downloads themes for a remote target that he could use themes on e.g. the Clip page as RWPS when renaming theme.wps to theme.rwps
13:57:18pixelmaI know it's complicated (and said so in the beginning), just a wish
13:59:01*pixelma tries to add a "get a notice" somewhere to her second last statement
13:59:52domonokymaybe just provide hints for uploaders when they upload a theme. IE "This theme contains a .rwps which would work as a main-wps on the follwing targets: .. if you repackage and reupload the remote skin files as main-skin files"
14:00:50domonokydoing this magic on the server gets complicated, as there is also .sbs/.rsbs and maybe in the future even more skin files.
14:01:21pixelmayes, that's what I meant - and another hint for downloaders where else he could look for RWPSs (and that he needs to rename the file - and needs to look for font sizes)
14:01:51domonokyonly that i would provide this hint to uploaders, not downloaders :-)
14:02:09domonokydownloaders should just be able to download a zip and have it working.
14:04:32pixelmawell, I think it would need both because I don't think someone who made a theme for the Clip would like to make a full blown theme for the H300 just because his version would work on its remote - the other way would then be to have rwps-only themes on remote target pages
14:04:33***Saving seen data "./dancer.seen"
14:07:38 Quit Grahack (Connection timed out)
14:08:07 Quit petur ("out for some shopping and fresh air")
14:09:42 Join Grahack [0] (
14:09:42archivatorn1s: did you get a chance to look at FS #10816 ?
14:09:51domonokyhm, true using a main theme as remote theme is difficult, because its not a complete theme. And you can really combine a remote theme and different main.theme easily (because of font sizes, icons, colors, etc). But the other way round would easy todo for uploaders to repackage.
14:10:51 Quit flydutch ("/* empty */")
14:11:24n1sarchivator: ah, no forgot about that and it isn't my area but i'll look if i find time and boredom :)
14:11:58archivatorn1s: that's okay. I'm almost done with the spectrum analyzer anyway :)
14:13:30n1slooks like a nice plugin :)
14:13:56n1swe sure do have enough bugs in rockbox and i like looking at ones i think i can fix :)
14:17:32 Join liar [0] (n=liar@
14:17:53archivatorI know it worked the last time I posted a patch and doesn't now. That's a ~6 month window - should be relatively easy to do a binary search for the bug :)
14:20:19 Quit teru ("Quit")
14:20:24Unhelpfuln1s: if you haven't tested, don't bother. i simulated the cf shifter behavior on arm, mod 256 shifts on arm are fine but mod 64 shifts overflow too soon.
14:21:01 Join teru [0] (
14:21:20 Quit gtkspert_ (Read error: 60 (Operation timed out))
14:22:41 Join gtkspert [0] (
14:22:52n1sUnhelpful: ah, the test is running (and can't be cancelled)
14:23:09 Join Thundercloud [0] (
14:23:35n1sand omfg aache is sloooow on coldfire
14:24:56n1syour latest improvement sped it up 0.5% on the sample in test_files
14:26:15Unhelpfulthat matches what i saw on e200 :)
14:26:51n1sstill around 33% realtime though so we need some kind of revolution to be able to get this realtime
14:30:09n1si don't understand your arm specific c code but if it relies on shift by negative isn't that undefined in c?
14:30:38 Join domonoky1 [0] (
14:31:09n1sbut a smart way to shift one way for positive counts and the other for negative could be useful i other places too, like tremor
14:32:21 Quit Grahack (Read error: 110 (Connection timed out))
14:33:01 Quit domonoky (Read error: 104 (Connection reset by peer))
14:33:49 Join GeekShadow [0] (n=Antoine@reactos/tester/GeekShadow)
14:34:17n1syeah, crc changed when using that c code on coldfire
14:34:30 Join Grahack [0] (
14:38:01 Quit tg1 (Remote closed the connection)
14:38:17 Join efyx_ [0] (
14:39:36 Join tg1 [0] (n=thomas@
14:40:16Unhelpfuln1s: it's undefined by C spec, but the behavior of the arm shifter is defined quite clearly. :)
14:41:06Unhelpfulregister shifts are by (shift & 255), so a negative register value is a positive shift of greater than the register width... if the magnitude of the negative value is small enough
14:41:16n1si hope we can rely on gcc not deciding to go crazy just because it can then :)
14:41:21Unhelpfulcoldfire is similar but has a different definition of "small enough"
14:41:52 Join kugel [0] (n=kugel@rockbox/developer/kugel)
14:42:13Unhelpfulgcc's behavior is probably pretty well-defined here, too... it will emit a shift operation, or roll the shift into the argument of another operation if possible.
14:42:16n1show big is small enough?
14:44:46n1sso the win is that yo do the shift and test after that so if the first shift is correct you saved a cmp? or am i lost again?
14:45:40kugelUnhelpful: a comment when exploiting undefined behavior doesn't hurt. I was wondering why plain C code is only used for ARM when looking at the commit
14:53:14 Join petur [0] (
14:59:59 Quit gtkspert (Read error: 110 (Connection timed out))
15:07:19 Quit Zarggg (Read error: 110 (Connection timed out))
15:12:18 Quit w1ll14m (Read error: 54 (Connection reset by peer))
15:13:57 Quit MethoS- (Remote closed the connection)
15:17:59 Quit Slasheri (Read error: 148 (No route to host))
15:21:07 Quit teru ("Quit")
15:21:30 Join Zarggg [0] (
15:22:46 Quit Grahack (Read error: 110 (Connection timed out))
15:24:19 Join phanboy4 [0] (
15:24:38 Join Grahack [0] (
15:32:19Unhelpfuln1s: negative shifts with values > -225 will have 0 output on arm
15:32:53 Quit Grahack ("Leaving.")
15:33:01 Join Lear_ [0] (i=chatzill@rockbox/developer/lear)
15:33:44Unhelpfulon coldfire only the low 6 bits are used for shifts, so that overflow happens at -33
15:33:56 Quit Lear_ (Read error: 104 (Connection reset by peer))
15:34:06Unhelpful63&-33 == 31
15:34:21 Join Lear_ [0] (i=chatzill@rockbox/developer/lear)
15:38:07n1sah i think i get it then
15:39:16n1sbut yeah, a comment would be nice :)
15:40:21 Join bimbel [0] (n=Miranda@unaffiliated/bmbl)
15:41:31Unhelpfuln1s: as far as why this wins goes, gcc just acts a bit smarter with that version. the macro causes gcc to emit a cmp instruction to test the sign of the shift, and then to follow it with a series of instructions with gt or le conditions. it could save *an* instruction here by using rsbs to do the subtraction and set flags at the same time, i'm not sure why it doesn't. it also rather brilliantly calculates shift *again* for each shift done.
15:41:31Unhelpful might be made better just by using an inline function, gcc sometimes fails to do good CSE.
15:43:15Unhelpfulso it calculates shift *three* times, and one of those is only used to test the sign, and then discarded.
15:44:13n1sUnhelpful: is that 4.0 or 4.4?
15:44:26Unhelpfulthat's 4.4
15:44:54n1syeah, that sounds braindead indeed
15:45:32n1sis it as stupid at higher O levels?
15:45:49n1sO1 is used for arm libfaad
15:50:45 Quit Lear (Read error: 110 (Connection timed out))
15:50:56 Join w1ll14m [0] (
15:51:08CIA-6New commit by Domonoky (r23972): show more info on admin checkwps page.
15:51:30 Quit archivator ("Leaving")
15:51:33Unhelpful-O2 doesn't do any differently. -O3 produces the same bad code but reorders instructions more aggressively.
15:53:13 Quit archivator_ (Remote closed the connection)
15:54:30bluebrotherdomonoky1: while you're at it, the themes installation is currently broken in svn rbutil due to the changed target names. Do you think there's an easy way to allow both to work for now?
15:55:56domonoky1bluebrother: does rbutil now use the newnames internally ?
15:56:17bluebrotherdomonoky1: yes
15:56:34 Quit Tomis ()
15:57:11Unhelpfulgcc 4.0 generates essentially the same code as 4.4 for all of these situations, but is slightly less clever about register allocation, so that the result is computed in r3 and needs to be moved to r0 before returning. inlining pow2 would fix this since the calculations using the result would just continue with with using that register as the result. using inline asm for DESCALE_SHIFT would probably *not* solve this problem on its own.
15:57:14n1sheh, O2 on libfaad slows decoding regular aac down by about 5% on coldfire :)
15:57:44domonoky1hm, so we should change the checkwps names and the target names on the themepage too.. but that will break current released rbutil.
15:59:20Unhelpfulrewriting the logic of the DESCALE_SHIFT macro as an inline function produces *slightly* better code than the macro... should probably make that the "generic" version of the function, and comment the arm behavior on which that version relies.
15:59:31 Quit Highlander ("Quitte")
16:00:16 Join bertrik [0] (
16:00:22CIA-6New commit by nls (r23973): Improved bitrev12, using a larger LUT to save one lookup, speeds up vorbis about 2% on beast, about 0.5% on h300
16:02:51Unhelpfulheh, beast wins most, yet again :/
16:03:15n1syes, probably some cache thing
16:04:04n1sthis will likely just be intermittent anyway as it seems work is underway to replace the whole mdct
16:04:28n1si wonder if faad is using iram optimally...
16:04:35***Saving seen data "./dancer.seen"
16:05:35n1si can't imagine it being bad for pp either but probably closer to coldfire
16:06:58 Quit phanboy4 ("Leaving")
16:07:07Unhelpfuldoesn't coldfire have serious issues for data-in-sdram?
16:07:32n1sit has no dcache at all so it's slow as hell basically
16:08:38n1shmm, faad is only using iram for constants and buffers, no code, i wonder how much is unused
16:12:21n1sabout 44.7kB is used so we should have about 3.3kB left, any idea which function is the hottest?
16:24:47 Quit BlakeJohnson86 ("Leaving.")
16:30:32 Join Tomis [0] (n=Tomis@
16:32:13 Join Slasheri [0] (i=miipekk@rockbox/developer/Slasheri)
16:32:26 Quit tg1 ("Leaving.")
16:39:48 Join tomers [0] (
16:43:20 Join evilnick [0] (i=4d63ba66@rockbox/staff/evilnick)
16:43:38evilnickgevaerts: Pingitty-pingitty-ping
16:44:36evilnickWrong channel, sorry
16:45:01 Quit bimbel ("Woah!")
16:45:06 Join froggyman [0] (
16:47:52 Quit Tomis ()
16:52:10 Join Tomis [0] (n=Tomis@
16:52:40 Quit tomers ("ChatZilla 0.9.85 [Firefox 3.5.5/20091109125225]")
16:55:18 Quit liar (Read error: 113 (No route to host))
16:58:36n1si wonder why libffad is using uint8_t as index and loop variables in many places...
17:00:51 Join robin0800 [0] (
17:06:21 Quit froggyman (Read error: 104 (Connection reset by peer))
17:12:46Unhelpfuln1s: not sure. is it likely to matter? i would *hope* that if gcc needs to spill such a variable from register to stack that it will just save the whole register...
17:13:34 Quit stoffel (Read error: 113 (No route to host))
17:13:35n1syeah, i hope so but is looks weird
17:15:06Unhelpfulis there an icache on coldfire targets? do we gain anything by moving hot code into iram?
17:20:36 Nick fxb__ is now known as fxb (
17:20:44Lear_Unhelpful: Moving code helps, but not that much.
17:21:13 Quit stooo ("Leaving.")
17:22:18Unhelpfulso instructions are cached, or the memory controller issue streams reads on them, at least?
17:23:15Lear_Don't know really.
17:24:14Unhelpfuli wonder if valgrind could be used on the sim to identify hot data... might help to tell us which tables or buffers most need to go in iram
17:31:59 Join BlakeJohnson86 [0] (
17:35:17 Join DerPapst [0] (
17:37:34 Join anewuser [0] (i=anewuser@unaffiliated/anewuser)
17:41:42n1sUnhelpful: yeah they have a small icache
17:42:41 Join stoffel [0] (
17:43:55 Join liar [0] (n=liar@
17:46:16 Join Tomis2 [0] (n=Tomis@
17:47:09 Join Lss [0] (
17:51:33 Join perfectdrug [0] (
17:52:18perfectdrughi, why are some manual links missing frome the manual page? e.g. samsung players show a picture but no links
17:52:47kugelthere's no manual for the samsung players
17:52:59kugelI think i have told zagor about it, but I may have not
17:53:05perfectdrugwhy is the picture for it there?
17:53:12perfectdrugah ok
17:53:30kugelbecause some internal target name management changed
17:53:32perfectdrugalso the yh-820 is a foto and not based og the svg
17:54:04kugela good photo apparently :)
17:54:38kugelhaven't you made some SVGs?
17:54:52perfectdrugyes its in svn already
17:55:53perfectdrugI have a little sparetime right now, any objections against a svg for the onda vx747 for the themes page?;) pixelma?
17:57:58perfectdrugi hope i get the orientation right, the vx747 is portrait not landscape right?
17:58:07 Quit Thundercloud (Read error: 54 (Connection reset by peer))
17:58:52 Join Tomis2_ [0] (n=Tomis@
17:58:56kugelperfectdrug: no objections at all ;)
17:59:25kugelthe sim indicates its portrait, I don't know if it could be flipped
17:59:55perfectdrugok so I try, the shiny metal border could be a little tricky though
18:00:24 Join Thundercloud [0] (
18:02:08CIA-6New commit by Domonoky (r23974): Allow users to update a theme if themename, lcdsize and author matches. Old theme gets hidden on emailverification. Also allow dublicate themenames if ...
18:03:46kugeldomonoky1: email doesn't need to match?
18:04:00kugelanyway, awesome \o/
18:04:03domonoky1kugel: email has to match too.
18:04:06 Join ajb [0] (
18:04:11 Quit Tomis (Read error: 110 (Connection timed out))
18:04:15domonoky1its now live :-)
18:04:36***Saving seen data "./dancer.seen"
18:05:08ajbWho's the best person to ask about looking at the playlist code? I've had a few comments but no one seems comfortable reviewing the code
18:06:11 Quit Tomis2_ (Read error: 60 (Operation timed out))
18:06:37n1sajb: unfortunately noone around now seems very familiar with it
18:07:17kugeldomonoky1: is the update process described somewhere or is it a hidden feature?
18:07:42ajbn1s: Oh well, I shall leave it pootling along then...
18:08:08 Quit Tomis2 (Read error: 110 (Connection timed out))
18:08:22gevaertsajb: this could be an opportunity for you to become that person! ;)
18:08:58ajbgevaerts: Sure, but I could hardly review my own code, of course it's fine :-)
18:09:01domonoky1kugel: pretty hidden at moment :-) if you happen to enter the correct data, the errormessage for dublicate themename will change, and offer a checkbox to update the old theme :-)
18:10:03ajbWell the last people to touch the apps/playlist* code where stripwax, n1s and kugel
18:10:25n1sajb: yes?
18:10:57*kugel declines to have any knowledge
18:11:39ajbn1s: I think you last change was various casting fixes (trunk@22440)
18:12:42 Join archivator [0] (n=archivat@
18:12:43 Quit Lear_ (Read error: 104 (Connection reset by peer))
18:13:04 Join Lear [0] (i=chatzill@rockbox/developer/lear)
18:13:44n1sajb: that rev had nothing to do with playlists, just some callbacks used by the list code
18:14:40ajbn1s: Sure, I'm just looking to the logs for potential reviewers. I think stripwax is the person that made the most changes to the logic
18:14:59ajbI don't suppose I could convince anyway to test the patch?
18:15:07n1sif you check the logs you also see *what* was changed
18:15:18n1si don't have time now
18:15:29*ajb switches to git log -p
18:24:41 Quit Thundercloud (Remote closed the connection)
18:30:57 Join aaron424 [0] (
18:31:02 Join fdinel [0] (
18:41:53 Join Tomis [0] (n=Tomis@
18:43:41n1sUnhelpful: using bs_generic for speex gives about 0.1% speedup with the LUT in iram so i guess it's not really worth wasting much effort on...
18:45:12 Join panni_ [0] (
18:48:28 Quit grndslm (Client Quit)
18:48:40 Quit Lear ("ChatZilla 0.9.85 [Firefox 3.6b4/20091124213835]")
18:53:37 Nick krazykit` is now known as krazykit (
18:54:45 Quit robin0800 (Remote closed the connection)
18:55:23 Quit Tomis (Read error: 104 (Connection reset by peer))
18:56:49 Join grndslm [0] (
18:58:26 Join Grahack [0] (
19:01:12 Join Llorean [0] (
19:02:24 Join Zaba_ [0] (n=zaba@about/goats/billygoat/zaba)
19:02:30 Join Tomis [0] (n=Tomis@
19:07:37 Quit Zaba (Read error: 110 (Connection timed out))
19:08:49 Join Tomis2 [0] (n=Tomis@
19:08:57 Quit w1ll14m (Read error: 110 (Connection timed out))
19:12:08 Quit Tomis (Read error: 145 (Connection timed out))
19:13:30 Join Omlet05 [0] (
19:13:39 Quit Omlet ("( :: NoNameScript 4.22 :: )")
19:17:36 Nick Ypsy is now known as YPSY (
19:19:26 Join Tomis [0] (n=Tomis@
19:21:19 Join funman [0] (n=fun@rockbox/developer/funman)
19:23:47 Join kugel_ [0] (
19:24:56 Quit bertrik ("De groeten")
19:25:26 Quit kugel (Read error: 104 (Connection reset by peer))
19:26:14 Quit Tomis2 (Read error: 110 (Connection timed out))
19:27:17 Join Thundercloud [0] (
19:29:05 Quit amiconn (Nick collision from services.)
19:29:09 Join amiconn_ [0] (i=quassel@rockbox/developer/amiconn)
19:29:16 Nick amiconn_ is now known as amiconn (i=quassel@rockbox/developer/amiconn)
19:31:04 Join pamaury_ [0] (
19:31:26 Quit pixelma (Nick collision from services.)
19:31:27 Join pixelma_ [0] (i=quassel@rockbox/staff/pixelma)
19:31:48 Nick pixelma_ is now known as pixelma (i=quassel@rockbox/staff/pixelma)
19:34:03 Join Tomis2 [0] (n=Tomis@
19:35:24 Join flydutch [0] (
19:40:49 Quit Tomis (Read error: 110 (Connection timed out))
19:40:50 Nick Tomis2 is now known as Tomis (n=Tomis@
19:42:12 Quit pamaury (Read error: 110 (Connection timed out))
19:44:48 Quit petur ("later")
19:49:00 Quit stoffel (Read error: 60 (Operation timed out))
19:51:35 Quit perfectdrug ("CGI:IRC (EOF)")
19:59:23 Quit BlakeJohnson86 ("Leaving.")
20:04:38***Saving seen data "./dancer.seen"
20:06:12 Quit flydutch ("/* empty */")
20:06:33 Join BlakeJohnson86 [0] (
20:10:47 Quit aaron424 (Remote closed the connection)
20:10:59 Join Melena [0] (
20:12:58Melenahey guys :) i have a little problem..... my sansa reboots into the original firmware, when i plug in a usb-cable. is there a was to force the bootloader to boot rockbox?^^
20:13:31Melenaits a sansa e200( v2)
20:14:54 Join liar_ [0] (n=liar@
20:16:03funmanMelena: keep the center button pressed when plugging usb (you'll only be able to charge the e200 though)
20:16:07 Quit funman ("free(random());")
20:17:33Melena... i tried this before but id didnt work..... anyways... cya
20:17:38 Quit liar (Read error: 104 (Connection reset by peer))
20:17:46 Part Melena
20:23:31 Nick liar_ is now known as liar (n=liar@
20:26:19 Quit Horschti ("Verlassend")
20:29:40 Join GeekShado_ [0] (n=Antoine@
20:30:27 Join GeekShad__ [0] (
20:33:03 Join Horscht [0] (n=Horscht2@xbmc/user/horscht)
20:37:32 Quit GeekShadow (Read error: 145 (Connection timed out))
20:45:51 Join solexx [0] (
20:48:10 Quit archivator ("Leaving")
20:49:07 Quit GeekShado_ (Read error: 110 (Connection timed out))
20:50:37 Quit solexx_ (Connection reset by peer)
20:58:23 Join stoffel [0] (
20:59:32 Join phanboy4 [0] (
21:02:17 Join mc2739_ [0] (n=mc2739@rockbox/developer/mc2739)
21:06:26 Quit anewuser (Client Quit)
21:13:55 Join mc2739__ [0] (
21:13:55 Quit mc2739_ (Read error: 104 (Connection reset by peer))
21:20:30 Quit krazykit ("Connection reset by beer")
21:22:48 Join krazykit [0] (
21:24:09 Quit evilnick ("Page closed")
21:26:14CIA-6New commit by tomers (r23975): PictureFlow: Add PF_MENU action to touchscreen
21:27:20 Quit mc2739 (Read error: 110 (Connection timed out))
21:27:57CIA-6New commit by tomers (r23976): PictureFlow: Add PF_MENU action to touchscreen
21:28:18 Join tomers [0] (
21:28:36tomersAnyway to change r23975's log message? I've named it incorrectly... :-(
21:28:56amiconnn1s: In case your test file uses the qmf - that filter can be sped up significantly on coldfire
21:29:47 Quit Grahack ("Tu m'as vu ?")
21:31:03 Join mc2739 [0] (n=mc2739@rockbox/developer/mc2739)
21:31:09 Quit fdinel ("Miranda IM! Smaller, Faster, Easier.")
21:33:10n1samiconn: it's the test sample we host on the download server, i have no idea if it uses qmf
21:34:25 Quit GeekShad__ ("The cake is a lie !")
21:37:57CIA-6New commit by FlynDice (r23977): Sansa AMS: Implement 4 bit widebus mode for SD cards. ...
21:38:41kugel_FlynDice: maybe you want to work on fs#10805?
21:38:57FlynDice what's that?...
21:41:07amiconnn1s: Btw, thanks for r23969(+970)
21:42:23 Quit mc2739__ (Read error: 110 (Connection timed out))
21:42:42 Quit tomers ("ChatZilla 0.9.85 [Firefox 3.5.5/20091109125225]")
21:43:08FlynDicekugel_: I'll take a look later today, was up working all night and need a bit more nap time ;-)
21:44:05kugel_FlynDice: no hurry, I'm just saying because you're more likely to work on it soon. especially since my fuze broke :)
21:44:21 Quit stoffel (Remote closed the connection)
21:44:55n1samiconn: no problem :) fixing bugs is fun if they can be reproduced easily and fixed relatively cleanly
21:45:54 Join petur [0] (
21:45:54*amiconn likes if fixes like this (which actually adds a missing piece of code) even save some binsize
21:46:43FlynDicekugel_: I thought you stopped throwing them in the street after the last 1 broke ;)
21:47:04kugel_uhm...yea, I did :)
21:47:23*kugel_ explains again that it wasn't his fault
21:48:38CIA-6New commit by Domonoky (r23978): add rating to themesite. ...
21:50:34*domonoky1 made theme ratings live :-)
21:50:42JdGordondomonoky1: something is broken...
21:50:53JdGordonyou got rid of the works with latest release line?
21:51:23JdGordonalso.. why 10 rating levels? 1-5 is moer than enough
21:51:48*domonoky1 can still see the "Works with.. " lines.
21:52:18JdGordonhaha wierd.. all the mini2g themes ont work with 3.4 apparently
21:53:04domonoky1can you point me to such a theme ?
21:53:21JdGordoncan you force the rating stuff to the bottom of the cell to make it loo a bit better?
21:53:45JdGordonoh never mind.. i had the nano2g not mini2g site open... no 3.4 for it
21:54:12domonoky1JdGordon: feel free to improve the templace its in svn :-) (private/template/themeslist.tpl and theme.tpl :-)
21:54:35domonoky1yes, there is no release checkwps for nano2g
21:55:24domonoky1or if you can tell me exactly how it should look like.. i could do it too :-)
21:56:26JdGordonumm... either forced to the bottom of the cell, so themes with different description lengths all have tge rating section aligned
21:56:37JdGordonor, next to the rating stars..
21:56:51domonoky1ah, now i understand.. i will try.
21:57:00JdGordonor even better.. make the 5 stars clickable...
21:57:47CIA-6New commit by tomers (r23979): Bubbles: Remove unused keymaps
21:58:03 Join ansuz [0] (
22:00:26domonoky1JdGordon: that would be nice, but i dont wanted to add javascript thingis.
22:00:47JdGordoncourse you do! :D
22:04:03 Quit jfc^2 (Read error: 54 (Connection reset by peer))
22:04:17CIA-6New commit by tomers (r23980): Bubbles: Rename defines
22:04:26JdGordonanywho... very cool additions :) /me disappears
22:04:39***Saving seen data "./dancer.seen"
22:05:37bluebrotherdomonoky1: now make rating use AJAX and work by clicking the stars ;-)
22:06:13domonoky1bluebrother: same answer as to JdGordon, if you do the javascript part... :-)
22:08:36 Join saratoga [0] (i=9803c6dd@gateway/web/freenode/x-lhiszqxalazrulvl)
22:08:50 Join marga_ [0] (n=marga@
22:09:54marga_Hi! I totally screwed up (due to stupidly doing stuff when being too tired) :-\. I deleted the partition table from a Sansa View, and now -obviously- it won't boot at all. Is there ANYTHING that can be done?
22:09:55saratogaamiconn, n1s: for what its worth, the "64kaache.m4a" test file uses the qmf (its an SBR only feature), while the "nero_" samples do not (they're LC)
22:10:18saratogamarga_: i don't think anyone here is going to be able to help you with that
22:10:30saratogathis is a rockbox channel, not a sandisk one
22:10:46marga_saratoga: ok, re-routing me to the correct channel is acceptable, I couldn't figure out what to ask.
22:11:23marga_saratoga: I know, but I found lot's of info on the hardware in the rockbox forums, and thus thought maybe somebody knew how to help me, or at least where I could ask.
22:12:07marga_Sorry for the stupid mistakes, I'm still too tired to think.
22:12:14 Join p3tur [0] (
22:12:45gevaertsmarga_: sandisk has its own forums. I don't think many rockbox people actually know much about the view...
22:13:16marga_gevaerts: ok, thanks anyway
22:13:39 Quit petur (Read error: 104 (Connection reset by peer))
22:14:01CIA-6New commit by Domonoky (r23981): correctly display release version on single themepage.
22:14:31 Join robin0800 [0] (
22:15:53 Nick p3tur is now known as petur (n=peter@rockbox/developer/petur)
22:19:24n1ssaratoga: ok, good to know, maybe i'll look at that if bored in the future, are any profile results available for the aac codec? (and is this qmf thing subject to being replaced anytime soon?)
22:20:20saratogan1s: stripwax tried to profile aac w/sbr but it kept crashing on him
22:20:30saratogai don't know if theres ones for LC available
22:20:58saratogastripwax might have one, i can't remember
22:21:01n1sok, i guess i'll try to do that if i will try to optimize anything
22:21:11saratogai'm not planning to look at the sbr stuff in the near future, but i want to eventually
22:21:23saratogamaybe try and combine it with the stuff in mp3, mpc, or atrac3
22:21:41n1si assume your todo list is as long as mine :)
22:22:28n1show is the new mdct coming along, i saw the initial commits but they give little indication to the state of it
22:22:58markunsaratoga: do you know if the sbr stuff in mp3pro is the same as in aac?
22:23:21saratogamarkun: i've never looked at an mp3pro decoder, but from what i understand its very similar
22:23:47saratogan1s: we've just been chipping away at it for a couple months now, but at least it can actually decode audio at the moment
22:25:05n1sah, so basically "just" optimization left?
22:25:09 Quit Thundercloud (Remote closed the connection)
22:26:57saratogan1s: that and code clean up
22:27:05saratogai still have to figure out how i want to expose it to all the codecs
22:27:29saratogathe old mdct had the advantage of not needing any init, and using the same size buffers for all transform sizes, the new one is going to be more complicated
22:27:37 Join Thundercloud [0] (
22:28:20n1sah, yeah that could be troublesome when trying to use iram
22:29:24saratogai think what i'll have it do is have a global init function, and then one per transform size init function where the codec passes in a pointer to a buffer
22:29:31saratogaso that each codec can decide to use IRAM or DRAM
22:30:30CIA-6New commit by bluebrother (r23982): Make lipo calls silent.
22:30:32n1ssounds reasonable, how are the memory requirements compared to the current one?
22:31:01saratogasomewhat worse
22:31:33saratogathe trig tables are the same size I think, but the pre/post rotation steps that convert an fft to an mdct will need different tables
22:31:48saratogafortunately they're accessed less often then the other trig tables
22:32:05bluebrotherlinuxstb: I'm considering to make all those *patchers Makefiles create proper dependencies (like rbspeex does) and move some stuff into a generic Makefile (probably in rbutil/ or rbutil/tools). Thoughts?
22:32:27n1ssaratoga: maybe it's time to review the codec/core split of iram
22:34:22 Join JdGordon| [0] (n=Miranda@nat/microsoft/x-htxrknrngmsbjhba)
22:34:50saratogayes I've wanted to do that for a while now, but core is not something i know a lot about
22:35:35saratogathe other annoying this is that the trig tables for the pre/post rotations use alpha = 2 * M_PI * (i + theta) / n for the cos/sin argument
22:35:42saratogawhere theta is some phase offset
22:36:11saratogaso that means larger tables aren't simply better sampled versions of small tables
22:36:15JdGordon|domonoky1: did you intend that the single theme page use full width?
22:36:19saratogaso you actually have to store 1 table per transform size
22:36:33saratogawhich is ok for most codecs, but kind of sucks for wma with its 5 different transform sizes
22:37:08 Join ded [0] (
22:37:40 Join jfc [0] (
22:37:54domonoky1JdGordon| what do you mean with full width ? it should look like the themelist but only with 1 theme.
22:37:54 Quit bmbl ("Bye!")
22:38:38JdGordon|the cell uses the full width of the browser window (minus the menu on the left of course) instead of only being 1/3rd of it like usually
22:39:50domonoky1JdGordon| which browser ?
22:40:21JdGordon|its fine on ie8
22:40:31domonoky1hm, in firefox it also looks normal.
22:40:49domonoky1lets blame google :-)
22:41:05JdGordon|fair enough
22:42:11robin0800domonoky1: its fine on linux with chrome beta
22:43:43dedHi all. I play a track in album A, then switch to playing album B. Then I pause and wait until idle poweroff. On reboot I hit "resume playback". It resumes with album A. I was expecting album B. I am not suggesting this is a bug but can't work out why my system is behaving like this. Can anyone help?
22:43:44 Quit Thundercloud (Remote closed the connection)
22:44:32dedI was on 3.4 but have just updated to today's release. ipod 5.5.
22:46:17 Join Thundercloud [0] (
22:47:11 Quit parafin (
22:47:11 Quit Torne (
22:47:11 Quit Tuplanolla (
22:47:11 Quit shodanX_ (
22:47:11 Quit blithe (
22:47:11 Quit jds2001 (
22:47:11 Quit Kopfgeldjaeger (
22:47:11 Quit dmb (
22:47:22 Quit rjg (Read error: 110 (Connection timed out))
22:48:39saratogahuh the sin and cos tables are basically just mirrors of one another
22:48:44saratogamaybe they can be compacted
22:48:47NJoinKopfgeldjaeger [0] (
22:49:01bluebrotherded: how are you playing album B? Are you queueing it from the context menu?
22:49:25NJoinblithe [0] (
22:49:56 Join Torne [0] (
22:49:59dedbluebrother: no, I am browsing to it via Files then hitting select on a track
22:50:03 Join shodanX [0] (
22:50:18NJoinTuplanolla [0] (n=jani@unaffiliated/tuplanolla)
22:50:25gevaertsded: can you try checking the filesystem on the ipod?
22:51:19dedgevaerts: Yes I'm on linux and have the ipod here. What exactly would you like to see the output of?
22:51:42gevaertsrun fsck.vfat on it
22:51:58dedgevaerts: OK.
22:52:14NJoinparafin [0] (
22:52:15gevaertsIn my experience, weird problems like this are often caused by something strange with the filesystem
22:53:35 Quit Torne (Killed by (Nick collision))
22:53:35NJoindmb [0] (n=Dmb@unaffiliated/dmb)
22:53:35NJoinTorne [0] (i=torne@rockbox/developer/Torne)
22:53:35NJoinshodanX_ [0] (
22:53:39 Join TorneWuff [0] (
22:53:45 Quit dmb (SendQ exceeded)
22:54:03dedgevaerts: Right. Do I unmount first and pass e.g. /dev/sdc2 as the argument?
22:54:12gevaertsded: yes to both
22:54:34 Join dmb [0] (n=Dmb@unaffiliated/dmb)
22:54:37 Quit Omlet05 ("( :: NoNameScript 4.22 :: )")
22:55:00 Quit shodanX_ (Connection reset by peer)
22:57:38saratogahuh it looks like trig tables in the new mdct can be reduced in size considerably, but to control error in this approach indexing would become more complicated, and we'd still probably only get about ~16 bits accuracy
22:59:27saratogai wonder why the sincos_lookup0 lookup table in the current mdct doesn't seem to have this problem
23:01:57saratogadigging into this the current mdct seems to omit the phase factor
23:03:07 Quit Torne (Read error: 111 (Connection refused))
23:03:52saratogadoes anyone understand the current mdct? i have a lot of questions about what its actually doing
23:06:34 Join MethoS- [0] (n=clemens@
23:09:28 Join fml [0] (
23:10:54fmlbluebrother: hello. You were right about \textsf vs. explicit usage of cmbright. When I first used \textsf (and didn't get the expected result) I forgot to add \textbf. And hence thought that chapter headings use a different font.
23:12:35 Join Casainho [0] (
23:13:28 Quit Thundercloud (Remote closed the connection)
23:25:26 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
23:29:07 Quit ajb (Remote closed the connection)
23:29:54 Quit fml ("CGI:IRC")
23:31:16Casainhook :-)
23:31:25Casainhoany advances on Mini2440?
23:31:47CasainhoI guess no...
23:31:48dedgevaerts: The problem is still there. fsck.vfat -a did find some things to correct, including "File size is 106 bytes, cluster chain length is 0 bytes." for /.rockbox/most-recent.bmark, but now it runs with no output.
23:32:05Casainho(sorry, mistake on channel)
23:36:13 Quit petur ("reboot :(((((")
23:36:15bluebrotherded: can you use chkdsk /f on windows? That sometimes finds issues fsck.vfat does not, at least from my experience.
23:36:52dedbluebrother: OK, will do.
23:38:39dedNo way! I just discovered you've made it control the volume on my computer while plugged on usb (even brings up the gnome notification graphic). Is there anything else I can do along those lines / is there some documentation available?
23:39:28AlexPyep :)
23:39:33AlexPIt should be in the manual
23:39:52AlexPYou can switch between e.g. multimedia, presentation (control powerpoint), mouse, etc modes
23:39:57 Join petur [0] (
23:42:28 Quit crwll (
23:43:09 Nick TorneWuff is now known as Torne (
23:43:52 Part ded ("ERC Version 5.3 (IRC client for Emacs)")
23:54:11 Quit domonoky1 (Read error: 104 (Connection reset by peer))
23:54:51 Join froggyman [0] (
23:56:54NJoincrwll [0] (
23:57:40 Quit JdGordon| ("Miranda IM! Smaller, Faster, Easier.")
23:58:20 Quit pamaury_ ("exit(*(int *)0 / 0);")

Previous day | Next day