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 2014-12-11

00:02:00 Quit AlexP (Remote host closed the connection)
00:06:19 Quit lebellium (Quit: ChatZilla 0.9.91 [Firefox 35.0/20141201162954])
01:08:37 Quit sakax (Quit: Leaving)
01:08:38foolsh[Franklin]: You've ask about tlsf before, is at anything biflib can't do?
01:09:32 Quit ender` (Quit: 𝘙𝘦𝘷𝘪𝘴𝘦𝘥 𝘭𝘰𝘨 𝘭𝘦𝘷𝘦𝘭𝘴 𝘱𝘳𝘰𝘱𝘰𝘴𝘢𝘭: "𝓯𝔂𝓲," "𝔴𝔱𝔣," 𝘢𝘯𝘥 "𝕠𝕞𝕘.")
01:16:05 Join AlexP [0] (~alex@rockbox/staff/AlexP)
01:17:56 Quit babylonlurker (Ping timeout: 240 seconds)
01:44:55***Saving seen data "./dancer.seen"
01:50:33 Quit froggyman (Ping timeout: 256 seconds)
01:52:16 Join froggyman [0] (~frogs@unaffiliated/froggyman)
01:56:19 Join chrisb [0] (
02:17:09 Join Strife89 [0] (
02:18:20 Quit shamus (Ping timeout: 250 seconds)
02:19:17 Join shamus [0] (
02:30:14[Franklin]foolsh: hadn't thought of it yet :)
02:30:32[Franklin]I don't need to free()
02:35:05[Franklin]but honestly buflib seems a bit overkill for what I want
02:39:19[Franklin]TBH, I could write my own "allocator" with a pointer and increment it each time :)
02:41:08[Franklin](this is for the string table loading)
02:43:12 Quit ZincAlloy (Quit: Leaving.)
02:43:35 Quit krabador (Quit: Sto andando via)
02:52:29 Quit derf (Ping timeout: 256 seconds)
02:54:15 Join derf [0] (
03:21:10 Join mikebeauc [0] (~mike@
03:22:49[Franklin]but really, how do I link with TLSF?
03:25:31 Join cmhobbs [0] (~cmhobbs@fsf/member/cmhobbs)
03:28:23 Quit [Franklin] (Remote host closed the connection)
03:44:57***Saving seen data "./dancer.seen"
03:47:55 Quit tchan (Quit: WeeChat 1.0.1)
03:54:46 Quit mikebeauc (Ping timeout: 265 seconds)
03:59:17 Quit preglow (*.net *.split)
03:59:17 Quit aevin (*.net *.split)
04:03:58 Join mikebeauc [0] (~mike@
04:04:54 Join preglow [0] (~thomj@2001:840:4243:3::100)
04:04:54 Join aevin [0] (eivindsy@unaffiliated/aevin)
04:08:41 Quit gevaerts (Ping timeout: 244 seconds)
04:09:27 Quit evilnick (Ping timeout: 272 seconds)
04:10:11 Quit AlexP (Remote host closed the connection)
04:13:01 Quit cmhobbs (Remote host closed the connection)
04:23:49 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
04:34:43 Join rela_ [0] (
04:35:32 Join evilnick [0] (~evilnick@rockbox/staff/evilnick)
04:36:34 Join gevaerts [0] (~fg@rockbox/developer/gevaerts)
04:38:35 Quit rela (Ping timeout: 272 seconds)
04:41:08 Join AlexP [0] (~alex@rockbox/staff/AlexP)
04:46:48 Join Strife1989 [0] (
04:48:36 Quit Strife89 (Ping timeout: 240 seconds)
05:22:14 Quit Provel (Read error: Connection reset by peer)
05:23:54 Quit TheSeven (Ping timeout: 244 seconds)
05:25:39 Join TheSeven [0] (~quassel@rockbox/developer/TheSeven)
05:27:12 Join Provel [0] (
05:44:58***Saving seen data "./dancer.seen"
05:53:12 Quit chrisb (Ping timeout: 265 seconds)
06:04:36 Quit AlexP (Remote host closed the connection)
06:48:51 Quit Guest66888 (Ping timeout: 244 seconds)
06:49:40 Quit charlie (Ping timeout: 272 seconds)
06:59:30 Join charlie [0] (~c@unaffiliated/charlie)
07:45:00***Saving seen data "./dancer.seen"
07:53:53 Quit mikebeauc (Quit: Leaving)
07:59:01 Quit mc2739 (Ping timeout: 250 seconds)
08:00:51 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
08:02:54 Quit preglow (*.net *.split)
08:02:54 Quit aevin (*.net *.split)
08:03:09 Join wodz [0] (
08:04:34 Quit Strife1989 (Ping timeout: 250 seconds)
08:06:35wodz[Franklin]: (log) look at lua.make or imageviewer/gif/gif.make (re looking how to use tlsf in plugin)
08:08:06 Join preglow [0] (~thomj@2001:840:4243:3::100)
08:08:06 Join aevin [0] (eivindsy@unaffiliated/aevin)
08:15:34 Join LinusN [0] (
08:26:55 Join ender` [0] (
08:32:29 Quit wodz (Quit: Leaving)
08:41:31 Quit mirak_ (Quit: Ex-Chat)
08:57:42 Join mortalis [0] (~kvirc@
09:17:46 Join einhirn [0] (
09:25:16 Quit [Saint] (Ping timeout: 240 seconds)
09:25:48 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
09:45:01***Saving seen data "./dancer.seen"
10:09:23 Quit pamaury (Ping timeout: 245 seconds)
10:12:50 Join [Saint] [0] (~saint@rockbox/staff/saint)
10:20:48 Quit mc2739 (Ping timeout: 255 seconds)
10:29:42 Quit Ivoah_ (Ping timeout: 264 seconds)
10:35:51 Join Ivoah [0] (
10:38:05 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
10:39:58 Join wodz [0] (
10:44:33 Quit mc2739 (Ping timeout: 250 seconds)
10:45:05 Join maruk1 [0] (
10:46:42 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
10:48:12 Quit Guest90896 (Read error: Connection reset by peer)
10:48:13 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
10:52:01pamaurywodz: ping
10:52:26wodzpamaury: pong
10:53:02pamauryI'm still thinking about the new register description format, I'd like your opinion, still on the problem of SCT
10:53:16pamaurytake this example: pamaury/36cffa3f8401ddd3f554">
10:53:37pamaurysay I want to have register ENABLEn with SCT variants, how would I add that to the format ?
10:54:37pamauryI see three possibilities: 1) duplicate SET,CLR,TOG variants [seems pretty bad] 2) introduce a new tag <variant> to put into the node alongside <register> 3) consider <node> under a <node> with <register> as variants
10:56:14wodz1) is not an option 2) breaks regularity
10:56:40wodzfor 3) what you mean by <register> as variants?
10:57:11pamaurylet me write an example
10:58:11 Quit mc2739 (Ping timeout: 245 seconds)
10:59:17pamauryI updated the gist
11:01:17pamauryI think 3) is the most natural although it seems convoluated at first, the idea is that it would define {COP_INT,CPU_INT(m)}.ENABLEn and {COP_INT,CPU_INT(m)}.ENABLEn.{SET,CLR} which both inherit the same register description
11:02:18wodz3) looks reasonable considering we abstract <node> so much
11:04:27wodzIts just a bit weird to follow this in xml BUT as we decided this approach targets machine processing and human readability is just bonus
11:05:17wodzthis will affect lua code, am I right?
11:07:47wodzpamaury: ^
11:07:51pamauryI cannot say right now, but it should not affect Lua so much, because the lua uses the "." syntax: COP_INT.ENABLEn.SET, so if the <node> hierarchies generates the same paths, the lua code will be unaffected. But the glue code in hwstub_shell will change for sure
11:08:48wodzah, ok. I thought that it uses DEV.ENABLEn_SET convention
11:09:15pamauryI'd need to check but I think at the moment it's .SET
11:10:21pamauryerr at the moment there is a .set() method which uses SET variant if available, and does RMW otherwise
11:10:48wodzpamaury: any thoughts about macros?
11:14:47pamauryyes, I thought maybe we could use BF_SET for what you suggest (Read-Modify-Write) and for each register R and field F in R, generate the following macros:
11:14:47pamaury- R_F(x) -> generates value where field F has value x (ie (x) << pos mostly)
11:14:47pamaury- R_F_V(val) -> same but for named values
11:14:47DBUGEnqueued KICK pamaury
11:14:47pamaurythis way the old BF_SET becomes R_SET = R_F(x) which okay in term of readability I think
11:17:00pamaurymaybe we could even generate a per-register macro for it:
11:17:00pamaury- R_F_WR(x) -> writes value x to the field
11:17:00pamaurythis would apply to normal and SCT variants, but for SCT it would be just a write, whereas for non-SCT it would be RMW
11:18:27pamauryfollowing this scheme you could even have R_F_SET()
11:20:04 Join AlexP [0] (~alex@rockbox/staff/AlexP)
11:22:00pamaurywhat do you think ? My idea is to keep the generic macros for multi-field accesses, but make special macros when you edit just one field, for convinience because this a very common case
11:32:39 Quit shmibs (Ping timeout: 250 seconds)
11:44:45 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
11:45:05***Saving seen data "./dancer.seen"
11:46:07 Join shmibs [0] (~shmibs@
11:59:59 Join edhelas [0] (~edhelas@2001:980:759b:1:863a:4bff:fe85:8a3c)
12:03:15wodzI don't like R_F() name (too cryptic) but the idea behind is good. This makes writing multiple fields at once easy
12:03:20wodzpamaury: ^
12:03:55 Quit shmibs (Ping timeout: 252 seconds)
12:13:45 Join shmibs [0] (~shmibs@
12:17:07 Quit einhirn (Ping timeout: 252 seconds)
12:21:13 Join xorly [0] (
12:24:50pamauryR_F is not a macro, a practical example would be: HW_ICOLL_ENABLEn_TYPE_V(NMI)
12:25:19pamauryand to write the field you would say: HW_ICOLL_ENABLEn_TYPE_WR_V(NMI)
12:25:32pamauryerr, I missing the value of n in the last example but you get the idea
12:25:33pamaurywodz: ^
12:27:37wodzah, so you plan to generate this for every register? Would it be readable at all? The names will be quite long.
12:28:30pamauryI can't see how you could shorten it: you have to specify anyway the register name, the field name and the value, this macro has each one exactly once
12:35:20 Join krabador [0] (~krabador@unaffiliated/krabador)
12:42:38 Join Guest66077 [0] (
12:45:09 Quit Guest66077 (Read error: Connection reset by peer)
12:47:25 Join TheLemonMan [0] (~lemonboy@unaffiliated/thelemonman)
12:50:38 Join Guest66077 [0] (
12:53:51 Join sakax [0] (~sakax@unaffiliated/sakax)
12:57:05 Join petur [0] (~petur@rockbox/developer/petur)
13:11:24wodzpamaury: yeah, that is probably unavoidable
13:12:20 Quit Guest66077 (Read error: Connection reset by peer)
13:16:05 Join Guest66077 [0] (
13:24:25 Quit krabador (Quit: Sto andando via)
13:32:23 Quit AlexP (Remote host closed the connection)
13:35:27 Quit xorly (Ping timeout: 244 seconds)
13:39:23 Quit edhelas (Ping timeout: 244 seconds)
13:45:08***Saving seen data "./dancer.seen"
14:05:35 Quit [Saint] (Ping timeout: 240 seconds)
14:30:58 Join AlexP [0] (~alex@rockbox/staff/AlexP)
14:31:41 Join meek [0] (981a1ad5@gateway/web/freenode/ip.
14:37:49 Join alucryd_ [0] (alucryd@archlinux/trusteduser/alucryd)
14:38:02 Quit alucryd (Read error: Connection reset by peer)
14:47:06 Quit meek (Ping timeout: 246 seconds)
14:52:58 Join einhirn [0] (
14:53:47 Join amayer [0] (
15:18:47 Part LinusN
15:24:06 Join edhelas [0] (~edhelas@
15:38:54 Quit wodz (Quit: Leaving)
15:45:10***Saving seen data "./dancer.seen"
15:49:55 Quit alucryd_ (Read error: Connection reset by peer)
15:49:58 Join alucryd [0] (alucryd@archlinux/trusteduser/alucryd)
16:06:11 Join ZincAlloy [0] (
16:32:15 Quit AlexP (Remote host closed the connection)
16:36:43 Quit Jinx (Quit: reboot)
16:45:35 Quit TheLemonMan (Remote host closed the connection)
16:47:48amayerDoes the newest patch for the ipod classic only effect video playback? or does it do something with how the screen is rendered also?
16:51:42prof_wolfffthe latest patch only affects YUV to RGB conversion, it is faster (more FPS in YUV mode) and should not affect other things
16:55:30amayerprof_wolfff: thanks. and thank you for all your updates to the ipod classic :)
17:03:18 Join Strife1989 [0] (
17:05:33prof_wolfffamayer: thanks, enjoy doing it
17:07:53 Quit Strife1989 (Ping timeout: 252 seconds)
17:33:53 Quit rela_ (Quit: Leaving)
17:35:59 Quit edhelas (Ping timeout: 260 seconds)
17:45:13***Saving seen data "./dancer.seen"
17:58:42 Join AlexP [0] (~alex@rockbox/staff/AlexP)
18:07:29 Quit mortalis (Remote host closed the connection)
18:08:22 Quit einhirn (Read error: Connection reset by peer)
18:30:34 Join xorly [0] (
18:32:13 Quit maruk1 (Quit: Leaving.)
18:55:09 Join edhelas [0] (
19:05:46 Join meek [0] (981a1ad5@gateway/web/freenode/ip.
19:05:56 Join bertrik [0] (~quassel@rockbox/developer/bertrik)
19:09:57 Quit meek (Ping timeout: 246 seconds)
19:12:28 Quit pamaury (Remote host closed the connection)
19:25:33 Join krabador [0] (~krabador@unaffiliated/krabador)
19:35:25 Join JdGordon [0] (
19:35:26 Quit JdGordon (Changing host)
19:35:26 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
19:37:04 Quit JdGordon_ (Ping timeout: 264 seconds)
19:40:26 Join JdGordon_ [0] (~jonno@rockbox/developer/JdGordon)
19:41:02 Quit JdGordon (Ping timeout: 244 seconds)
19:45:14***Saving seen data "./dancer.seen"
20:22:39 Join wodz [0] (
20:24:48 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
20:26:40 Quit JdGordon_ (Ping timeout: 272 seconds)
20:32:01 Join babylonlurker [0] (
20:41:27 Join przemhb [0] (
20:45:35wodzprof_wolfff: ping
20:51:06wodzcronix: how is your pono hacking going?
20:51:18cronixnot to bad
20:51:30cronixive dumped all the drives, rockbox worked
20:51:40cronixaswell as several other audio players :D
20:51:50cronixim waiting for the kernel sources now
20:52:16cronixwhen they arrive im gonna work on enabling the pono to act as an USB DAC
20:52:40cronixtheres a kernel driver which does exactly that
20:53:02wodzwhat was wrong, that rockbox didn't work?
20:53:18cronixpretty lame pebkac issue
20:53:35wodzthats common :-)
20:53:43cronixi put the rockbox apk into /system/app/ and flashed it together with the rom
20:53:50cronixthat usually works for alot of apps
20:53:55cronixnot for rockbox tough
20:54:15cronixonce i put a file manager into the rom and installed the apk via android it worked in an instant
20:54:52cronixthe guy who did the pono teardown blogpost contacted me
20:54:56wodzcronix: Would be nice if you cook some wiki page about all this
20:55:12cronixhes going to have an interview with the chief technology guy at pono
20:55:27cronixand ive poured some questions into the mix that im curious about
20:55:39cronixespecially about some debugging and rescue methods
20:56:04wodzrecovery method is probably the most important
20:56:45cronixrecovery and ADB
20:56:53cronixthose are my 2 main concerns atm
21:00:26cronixim currently wondering if i setup a whole wiki just for the pono
21:01:13wodzcronix: Maybe you don't know but we hava a tradition to post "have a sound" mails to ML when rockbox plays its first song on new platform.
21:01:17prof_wolfffwod: hi!
21:01:27prof_wolfffwodz: hi!
21:01:44cronixah okay i will write something up then real quick
21:02:08wodzprof_wolfff: I was looking at your freq scaling patch on gerrit. Why do you clock lcd so low in unboosted state?
21:02:22cronixeven with the android app? which is kind of a port itself
21:03:14wodzprof_wolfff: And as I wrote in a comment there should be test_codec results for unpatched version to have reference point and to judge what is reasonable freq for unboosted state
21:03:46wodzcronix: on the phone I wouldn't bother, on DAP which needs some hackery to run rockbox - why not
21:05:29cronixalright i will have something small written up in a few, can i just sent an email to some address or do i need to register with the mailing list?
21:06:02prof_wolfffLCD it is at maximum 'safe' frequency, it depends on APB, i need to do more tests to try to overclock LCD (it success on iPod 80 but fails on iPod 160) or increase APB
21:06:35prof_wolfffother way is that iPod target internally increases APB frequency when LCD is active (not sleeping)
21:06:58wodzI'd say increase APB. 9MHz is damn low for anything
21:07:50prof_wolfffyes LCD should run at least at the actual frequency
21:07:56wodzanyway without codec benchmarks I am hesitant to say that 36MHz is ok for unboosted
21:08:27prof_wolfffactually i have recent tests for actual SVN version
21:09:15cronixwodz: devel or users list?
21:09:25wodzand doing benchmarks for unboosted is not that trivial as AFAIK test_codec is always run boosted and recalculates result.
21:09:28wodzcronix, DEV
21:11:06prof_wolfffmy test are boosted but it is easy (but long) to to unboosted test modifying RB
21:11:09wodzprof_wolfff: commenting ATA driver change - AFAIK we always boost on rebuffering (which is 99.9% of when you access storage)
21:11:23cronixmailing is sent
21:12:44wodzprof_wolfff: If you don't change cpu/mem ratio I'd say you can safely extrapolate results and don't bother to tweak rb and conduct lengthy tests
21:13:29prof_wolfffi am not sure why i boost to read SMAT data, it is a very old patch that i updated, but i will see it
21:14:05wodzcronix: mail read :-)
21:14:27cronixGreat :)
21:15:11cronixreally the first new device in 2 1/2 years? :I
21:15:27wodzprof_wolfff: No I am commenting explicit cpu_boost(true) in freq scaling patch
21:15:44 Part przemhb
21:16:56 Join przemhb [0] (
21:17:14prof_wolfffok, will look at it, it is also an old patch update, i have used it for months and works well for mp3 and flac
21:17:49prof_wolfffanyways i will perform the unboosted codec_tests to see complete results
21:18:03wodzprof_wolfff: Could you add current HEAD results to CodecPerformanceComparison wiki?
21:18:54 Quit Guest66077 (Ping timeout: 244 seconds)
21:19:29wodzprof_wolfff: your DMA and UART patches look ok for me
21:20:17prof_wolfffwodz: nice! i think they need a bit more testing
21:23:50cronixall further progress depends on a working recovery method and a way of flashing new firmwares without the default player app
21:24:31cronixbasically what i want to do is tag the rockbox app as android launcher so android starts it when it boots and it feels more native
21:24:58cronixbut i cant do that if there is no other method of flashing since you cant run android apps from rockbox
21:25:20cronixand there is no way of going back to a other app once rockbox is defined as android launcher
21:25:38cronixso in the current state if i would do that one would be stuck with the installed firmware forever
21:25:53 Join TheLemonMan [0] (~lemonboy@unaffiliated/thelemonman)
21:27:19wodzI'd say write android luncher which lets you choose what to start. Timeout + default value will do. That is how we do on native hardware usually.
21:28:35cronixthat would be an option... but the default pono app is defined as launcher aswell
21:28:47prof_wolfffwodz: how can i add the test results to CodecPerformanceComparison? attaching test_codec_log.txt file?
21:29:01cronixso then we would get 2 selections of "boot-target" right after each other
21:29:12cronixandroid itself comes with a launcher selection
21:29:38cronixbut if anyone hits the "make this default" tickbox its stuck on said firmware version forever aswell
21:30:11 Join RiDD [0] (
21:30:34wodzcronix: custom luncher can react on button press or something - so when you boot holding some key down it will boot to OF
21:30:59wodzprof_wolfff: Adding entries in the table
21:31:15wodzprof_wolfff: Its manual process unfortunately
21:33:12cronixi will wait for sourcecode request answer and for the results of said interview befor i make any decisions
21:33:58cronixif the x-loader, uboot and kernel get properly released sourcecode we could even go more native, skipping the whole android layer
21:34:08prof_wolfffok, hope it will be not much dificult :)
21:34:29wodzprof_wolfff: It is always difficult to run on bare metal
21:34:58prof_wolfffwodz: :)
21:35:03wodzbah, I mean cronix
21:35:25cronixwell rockbox as android app misses a few functiosn the pono offers
21:35:36cronixfor example balanced output
21:35:42cronixswitching between jack modes
21:36:13cronix(2 headphones, 1 headphone + 1 line-out, 1 Balanced Output over both jacks)
21:36:52cronixstorage mode enable (to copy new tracks to the player) one has to drop to android to enable that
21:36:56cronixall not too handy
21:40:02 Join saratoga_ [0] (123e1f03@gateway/web/freenode/ip.
21:40:20cronixand besides, i want to implement this one so bad:
21:40:33cronixanyone tinkered with that module before?
21:40:45saratoga_prof_wolfff, wodz: test_codec can do both boost or unboosted tests (its an option in the menu)
21:41:55prof_wolfffsaratoga_: great, will try to launch it tonight
21:42:06saratoga_what is the concern with unboosted codec speed?
21:42:42saratoga_prof_wolfff, wodz: also there is a script in the rockbox utils folder (IIRC) that converts test codec logs to wiki tables
21:43:12saratoga_i think we have both a perl and ruby one
21:43:17prof_wolfffsaratoga_: that is much more great!
21:44:14saratoga_why did you want it though?
21:44:21saratoga_unboosted tests that is
21:44:34prof_wolfffi really don't know how RB decodes the audio, i was assuming the decode is boosted, i am right?
21:44:47 Quit shmibs (Ping timeout: 255 seconds)
21:45:15***Saving seen data "./dancer.seen"
21:45:42saratoga_it decodes into a PCM buffer unboosted, and if the buffer gets low, it boosts until its full
21:47:03prof_wolfffthen, it will be possible to decrease even more CPU speed
21:47:58saratoga_yes ideally the unboosted speed is relatively low
21:48:29saratoga_unless boosting is very costly, its best to frequently do it to avoid idle cycles
21:48:45saratoga_i'd recommend putting it just a little slower than is needed for mp3 decode
21:48:53saratoga_assuming you can
21:49:23prof_wolfffyes, i think we could reach 12MHz
21:49:30 Join lebellium [0] (
21:49:49saratoga_the GUI may be unusable that low
21:50:05wodzI bet UI performance will be unacceptable
21:50:33prof_wolfff12 MHz only playback, we can increase it (not completley boost) when the LCD is active
21:51:49prof_wolffflet's say the Classic port enters a superunboosted mode when there is only playback
21:52:11 Join [Saint] [0] (~saint@rockbox/staff/saint)
21:52:57saratoga_at very low clocks it may also be worth testing if you can undervolt the CPU more
21:54:02prof_wolfffprobably it could be undervolted
21:54:31saratoga_this is an arm9 not arm9E?
21:54:32wodzAFAIK the idea of multiple boost steps has been discussed several times and simple boost/unboost scheme works the best
21:55:11saratoga_varying the unboost clock based on LCD screen status may be reasonable, no idea how much difference it really makes
21:55:39prof_wolfffyes, i know, i am talking about a target specific mode, it is easy to decrease/increase freq when LCD sleeps/resumes
21:56:11saratoga_what arm processor is this?
21:58:15prof_wolfffwodz: about the cpu_boost(true), it is needed because HDD fails if HCLK decreases too much using current UDMA modes, it could work if we decrease DMA transfers speeds, but then the file reads takes more time and probably power consumption is higher than boosted HDD reads
21:58:27prof_wolfffit is arm926ej-s
21:58:29 Join shmibs [0] (~shmibs@
21:59:11saratoga_ok so same as amsv2
21:59:32saratoga_on those the cpu voltage can go very low if the clock is very low, and the savings is surprisingly large
21:59:49saratoga_perhaps this is more true for samsung parts given their fabs
22:00:08prof_wolfffyes, the 36MHz vs 54MHz is very noticeable
22:01:16wodzprof_wolfff: I think hdd access should be always boosted already. Thats why I am a bit surprised
22:01:50saratoga_given the cost of HD access its probably a good idea to boost unconditionally for it
22:04:33prof_wolfffthe cpu_freq patch is old, i am not sure how HDD power-up is actually done
22:04:48 Join Mihail [0] (2ebf0f36@gateway/web/freenode/ip.
22:06:12prof_wolfffbut i think there is no changes and actually HDD reads are unboosted
22:07:44Mihailfor clip zip I don't see big difference between 12 and 36 mHz
22:07:58Mihailit looks like cpu not consume very match
22:08:45prof_wolfffwodz: about the magic numbers, i also hate them :)
22:08:51Mihailwhen I play pcm it consume mostly same power like low bitrate mp3
22:13:16wodzprof_wolfff: Did you narrowed down the bug in ceata driver?
22:14:08prof_wolfffMihail: as saratoga_ said probably the limit is in between 12 and 36 MHz
22:15:07prof_wolfffwodz: are you referring to the ATA error?
22:16:24Mihailyes, 36 MHz good choose for GUI and power consumption
22:18:02prof_wolfffwodz: it is only for CEATA, it seems the panic is in ata_set_feature()
22:19:44prof_wolfffwodz: this functions was introduced for CEATA drives in the last driver patch
22:20:48Mihailprobably we can still have room for much lower power consumption - on pause zip consume 5ma (dac+output stage on), at playing pcm or low bitrate mp3 - 12-13 ma
22:23:13 Quit wodz (Quit: Leaving)
22:24:28prof_wolfffIIRC Classic playback @36MHz was about 16mA, but i cannot recall exactly the volume settings
22:26:07Mihailhow much it consume on pause?
22:27:04saratoga_theres actually some more mp3 optimizations i was hoping to finish one of these days
22:27:15saratoga_should reduce cpu cycles needed quite a bit
22:28:59Mihailimho it have litle sense as playing pcm consume to much
22:29:19prof_wolfffMihail: cannot remember that, it was not 5mA, probably ~10mA
22:30:41prof_wolfffMihail: i annotated all measurement data but need to look for it
22:32:10Mihailprobably zip have problem with dma (or don't idle on dma transfer)
22:33:47Mihailprof_wolfff: can you publish it?
22:34:22prof_wolfffthe measurements?
22:35:22prof_wolfffof course, will try to find and include them in the cpu freq patch info
22:51:30 Join mirak [0] (
22:53:35 Quit preglow (Quit: leaving)
22:55:24saratoga_could compare flac instead of pcm, its only a few MHz to decode and has lower bitrate
23:02:12 Quit edhelas (Ping timeout: 265 seconds)
23:03:38 Quit zoktar (Quit: -)
23:04:51Mihailflac - 12.5 ma, mp3(192) - 13.6, both test files from rockbox wiki
23:09:50 Join [Franklin] [0] (~franklin@unaffiliated/franklin)
23:09:56 Quit bluebrother (Ping timeout: 265 seconds)
23:11:06 Join zoktar [0] (~zoktar@unaffiliated/zoktar)
23:12:01 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
23:14:10 Join chrisb [0] (
23:14:46 Join edhelas [0] (
23:14:57 Quit edhelas (Remote host closed the connection)
23:21:28 Quit chrisb (Ping timeout: 272 seconds)
23:25:06 Join chrisb [0] (
23:29:37 Quit chrisb (Ping timeout: 244 seconds)
23:30:26 Join chrisb [0] (
23:41:56[Franklin]foolsh: I've gotten string table loaded re-implemented to allow for some more features
23:42:08[Franklin]I ended up not using tlsf
23:45:16***Saving seen data "./dancer.seen"
23:54:24 Quit amayer (Quit: Leaving)

Previous day | Next day