<Turl>
I fixed the android part with what I know and use
<Turl>
and I compared the hashes, it's not the same as any of the other r3p0s
mSquare has joined #arm-netbook
tekzilla has quit [Ping timeout: 264 seconds]
tekzilla has joined #arm-netbook
gsilvis_ has joined #arm-netbook
gsilvis has quit [Read error: Connection reset by peer]
grevaillot has quit [Read error: Operation timed out]
grevaillot has joined #arm-netbook
dfletcher_ has quit [Read error: Connection reset by peer]
dfletcher_ has joined #arm-netbook
uwe_ has quit [Read error: Operation timed out]
uwe_ has joined #arm-netbook
<Turl>
mnemoc: I tuned the captcha to use hand-set questions, let's see if this stops the spam registration :)
<Marex>
Turl: ;-)
<Marex>
Turl: do you now understand likely() and unlikely() ? :p
n6pfk has quit [Remote host closed the connection]
<lundman>
Unresolved reference: unlikey()
Quarx has joined #arm-netbook
vgrade has quit [Ping timeout: 268 seconds]
afgdsgh has joined #arm-netbook
rellla has joined #arm-netbook
<andoma>
anyone here that has exprience with opengl/mali using fb (ie, without x11)
mSquare has quit [Ping timeout: 246 seconds]
mSquare has joined #arm-netbook
<zub>
Marex: ex-fjfi, sadly
arete74 has quit [Ping timeout: 246 seconds]
arete74 has joined #arm-netbook
<mnemoc>
Marex: i thought it was a jump when true vs. jump when false thing
<mnemoc>
never looked into the generated asm
<zub>
hm, for the 1st time I connected a display to my mk802+, and tried to boot the stock android from flash. It's stuck in the "flashing android logo" screen (first some full-screen bitmap is displayed, then just "ANDROID" - the animation is running, but the screen stays on forever)
<zub>
I'd have though the stock system should boot. And as far as I can tell I have not touched the internal flash, so it should just work.
<mnemoc>
reflash it...
mysteryname has joined #arm-netbook
<lundman>
I've had it non-boot a few times, usually stuck on that garish mele screen animation. power cycle and it boots
sspiff has joined #arm-netbook
sspiff has quit [Changing host]
sspiff has joined #arm-netbook
mysteryname has quit [Ping timeout: 246 seconds]
afgdsgh has quit [Quit: ChatZilla 0.9.89-rdmsoft [XULRunner 1.9.0.17/2009122204]]
popolon has joined #arm-netbook
cat_n9 has joined #arm-netbook
afgdsgh has joined #arm-netbook
<oliv3r>
Hi, any awesome template creators available? (for the linux-sunxi wiki)
sspiff has quit [Remote host closed the connection]
<mnemoc>
we only have one template creator yet
sspiff has joined #arm-netbook
<mnemoc>
but it's pretty simple
sspiff has quit [Changing host]
sspiff has joined #arm-netbook
<oliv3r>
aye, but I don't have too much time, so either i keep going as I have, or write a template first
<oliv3r>
i think a template could be somewhat usefull to 'generate' the registers page from sub pages
<oliv3r>
omg, this HAS to be a 'bug' in the datasheet.
pawel5870 has joined #arm-netbook
* mnemoc
doesn't like the idea
<mnemoc>
for the initial content you can use an script to generate locally and paste the result
pawel5870 has quit [Ping timeout: 246 seconds]
<mnemoc>
but for later maintainance using templates for that becomes a limitation
<mnemoc>
for infoboxes and banners the templates work great. but not for content
<oliv3r>
AVS_CNT1_PS (pause the audio/video sync counter) is supposed to pause, and un-pause the AVS counter.
The-Compiler has quit [Ping timeout: 245 seconds]
<oliv3r>
but according to the datasheet, it's a Readonly value
The-Compiler has joined #arm-netbook
afgdsgh has quit [Quit: ChatZilla 0.9.89-rdmsoft [XULRunner 1.9.0.17/2009122204]]
<oliv3r>
well, all subpages list regiser addresses etc etc, so generating the table of registers, would always have up-to-date names, address and sizes
afgdsgh has joined #arm-netbook
<mnemoc>
rm: why not on linux-sunxi's wiki? :<
<rm>
because I :E a bit @ this cruel world
<rm>
where no one cares about a random newbie (in a particular field) and their problems
<mnemoc>
o.o
<mnemoc>
you asked about armhf, everyone told you it was just the same as any arm kernel...
<mnemoc>
and for the specifics about debian packaging i suppose no one knew
<mnemoc>
i'm not aware of any other debian maintainer here
<mnemoc>
(beside you)
pawel5870 has joined #arm-netbook
<rz2k>
Turl: thanks for wiki edit!
tzafrir has joined #arm-netbook
tzafrir_laptop has quit [Ping timeout: 248 seconds]
tzafrir has quit [Read error: Connection reset by peer]
tzafrir has joined #arm-netbook
<Marex>
zub: how come? They kicked you out ?
<Marex>
mnemoc: it tells the branch predictor if the result of the expression is true or false more often (and yes, it jumps based on the result)
<zub>
Marex: nope, finished
<Marex>
mnemoc: but obviously, the prefetch is also controller by the branch predictor
<Marex>
zub: good :)
<zub>
but no longer a student
<zub>
instead on the right track to grave :)
<Marex>
heh :)
<Marex>
zub: se you became misled from breaking the core to breaking the kernel ? :p
<zub>
rm: ad 'WTF': axp_power/* requires INPUT, and it's possible to unselect it and then end up with build failure. So I tried to add that. Tested with my config, but have not tested with defconfig/other configs.
<rm>
yeah, I get the point of your patch
<rm>
but somehow it fails with my config
<rm>
I will upload it a bit later
<zub>
Marex: 12:02 < Marex> zub: so you're trying to get upstream kernel running on some of the a10 devices ?
<zub>
well, sorry for breaking stuff; should have done more testing... :(
<zub>
Maybe I should have used depends on, rather than select. But select looked nicer.
QingPei has joined #arm-netbook
mSquare has quit [Ping timeout: 246 seconds]
<Marex>
zub: depends is a good pick, indeed
<zub>
Marex: but with depends it's easy to kill the option off without even seeing it; anyway select obviously doesn't work there
<mnemoc>
shouldn't menuconfig AW_AXP itself depend on INPUT?
<zub>
btw. I think I did test it with defconfig... not that it would matter much now
<Marex>
zub: you can use the "search" thingie to find it
<zub>
but why should be the input "global" for all AW_AXP?
<zub>
although... what is AX_AXP anyway? :)
<mnemoc>
just a holder for all AXP stuff
<mnemoc>
but if all choices depend on input...
akaizen has quit [Ping timeout: 260 seconds]
<zub>
currently yes, but does it have to be that way?
<zub>
anyway, yup, should work...
msb___ has quit [Ping timeout: 246 seconds]
msb___ has joined #arm-netbook
akaizen has joined #arm-netbook
<mnemoc>
Turl: the wiki broke :<
<oliv3r>
ok i'll copy my current section on the timer, since i haven't saved for two hours
<oliv3r>
when will the wiki back up?
<oliv3r>
also, feck. I just realized i used K for kilo, when I should have used k, but now i'm even more confused, becuase I thought in the case of KiB you do write it with a capital :S
<oliv3r>
or was it kiB?
<oliv3r>
ah, wikipedia knows, it is kB if you mean 10^3; but it is KiB if you mean 2^10th
<oliv3r>
why not make it conusing! I guess they wanted kibi to actually use a capital as the kilo should have been a capital imo :)
<mnemoc>
everyone sane understands that KB refers to 2^10
<oliv3r>
yeah, but I should have wrote it as kB :p
<oliv3r>
BUT! i don't think i used kilobyte anywhere, I think i used KiB properly :)
<oliv3r>
i DID goof it up on KHz though, but fixed that in my temporary document. just gotta wait for the wiki to reappear :)
<mnemoc>
:)
<oliv3r>
any eta?
<mnemoc>
oliv3r: for?
<oliv3r>
wiki to be back up
<mnemoc>
isn't it already?
<oliv3r>
lemme check
<oliv3r>
ah yes, preview works once more!
<mnemoc>
:)
<cat1>
mnemoc: any plan to push fcbe79e to linux-sunxi-3.4?
<mnemoc>
cat1: yes, absolutely. sorry... 2m
<cat1>
mnemoc: good, but no problem :)
<mnemoc>
last commit i accept when too sleepy to test myself :<
arete74 has quit [Read error: Connection reset by peer]
<mnemoc>
cat1: done
<cat1>
mnemoc: rebased, thanks!
QingPei has quit [Quit: Leaving.]
QingPei has joined #arm-netbook
QingPei has quit [Client Quit]
QingPei has joined #arm-netbook
QingPei has quit [Client Quit]
QingPei has joined #arm-netbook
QingPei has quit [Client Quit]
tzafrir has joined #arm-netbook
<mnemoc>
cat1: btw, did you get wifi working in 3.4?
<oliv3r>
anybody have any idea what 'GSM' could stand for in relation with the 32kHz timing crystal?
<mnemoc>
oliv3r: please hold, I'll mess with the wiki again
<oliv3r>
ok :)
<lundman>
Basic Cellular GSM/32 GSM/32.768kHz Clock General ?
ol1ver has joined #arm-netbook
<lundman>
seems standard GSM frequency
<oliv3r>
32kHz is THE default RTC freq.
<oliv3r>
anything that wants to accuratly keep track of time, has a 32kHz clock (there are alternatives of course)
<lundman>
yeah 31khz!
<lundman>
what, prime numbers are cool
<oliv3r>
but in case of the A10, it has a 24MHz oscilator for the chip itself, and uses the 32kHz to keep track of time etc
<oliv3r>
anyway, there is a register called EXT_LOSC_GSM that does 'something' GSM to it, you can set it to low, or high
<cat1>
mnemoc: same hack and behaviour as for/in 3.6 kernel: connection is rather unreliable, with random disconnections. I suspect power management issues, but did not really traced the problem further.
<mnemoc>
Turl: so now update.sh is only updating extensions
<mnemoc>
Turl: until they fix the updater script for pgsql
<Turl>
:)
mysteryname has joined #arm-netbook
<Turl>
mnemoc: btw, if you edit LocalSettings.php, towards the end, you can read the new captcha questions I added
<mnemoc>
is it only for registering?
<Turl>
for general captchas
<Turl>
if you fail to login repeatedly one pops up too I think
<mnemoc>
i hope it helps to reduce the account creating noise
<rz2k>
about that sata crap going on - we should just make CONFIG_ for RevC with this change and let it go. month for discussing a patch that people need is already too much.
<mnemoc>
i was going to apply it tonight unless someone how understand these clocks says the oposite
<mnemoc>
s/how/who/
<ibot>
mnemoc meant: i was going to apply it tonight unless someone who understand these clocks says the oposite
<mnemoc>
Turl: btw, does A10/JB work?
<rz2k>
Quarx has jb with cedarx working
<mnemoc>
decently?
<mnemoc>
i want to reflash the mele of my daughters
<Turl>
mnemoc: yeah it does
<Turl>
I started merging the cedar stuff from quarx the other day
<mnemoc>
so quarkx == AOSP, and you == CM?
<Turl>
mnemoc: no, Quarx == CM too
<Turl>
but (personally) his tree is unmanageable in the long run, as it breaks inter-device compatibility
<Turl>
so I merge his stuff on mine and add #ifdef on all the things
<Turl>
mnemoc: video decoding works
<Turl>
encoding crashes but it's a mele so you won't care :P
<Turl>
and camera doesn't work too well either
<mnemoc>
Turl: do you have a device tree for the mele?
<Turl>
nope
The-Compiler has quit [Quit: WeeChat 0.4.0-dev]
<mnemoc>
i thought you have gotten one
<mnemoc>
device i mean
<Turl>
yeah I do
<Turl>
I don't run android on it though
<mnemoc>
:)
<Turl>
mnemoc: btw, I sent a revised patch just after you left last night
drachensun has joined #arm-netbook
<mnemoc>
Turl: yes, saw it
<mnemoc>
Turl: still bothered by the likely() noise :p
<mnemoc>
but i'll merge it tonight
QingPei has left #arm-netbook [#arm-netbook]
The-Compiler has joined #arm-netbook
The-Compiler has quit [Quit: WeeChat 0.4.0-dev]
The-Compiler has joined #arm-netbook
The-Compiler has quit [Client Quit]
The-Compiler has joined #arm-netbook
The-Compiler has quit [Client Quit]
The-Compiler has joined #arm-netbook
<Turl>
mnemoc: I cleaned up most of wemac style :)
arete74 has joined #arm-netbook
<mnemoc>
Turl: hope you didn't delete all the "dead code"
<Turl>
mnemoc: no, I didn't
<mnemoc>
\o/
<mnemoc>
ML ;-)
<Turl>
I did however kill all of the // comments on it and replaced by #if 0
<mnemoc>
yes, that's good
The-C0mpiler has joined #arm-netbook
<mnemoc>
but until we are sure a piece of commented out code doesn't give us extra knowledge about the controller we can't wipe it out
<mnemoc>
that driver is everything we know about the wemac
<mnemoc>
if (1 || netif_running(ndev)) /* todo: ... */
<Turl>
-ETOOUGLYIMO
<Turl>
:P
<Marex>
zub: good thing is that the czech hacker crowd isn't shrinking ;-)
<Marex>
Turl: yep
<rm>
"1 || netif_running(ndev)" is kind of puzzling
<rm>
afaik some compilers evaluate from right to left
<mnemoc>
that would be against the standard
<rm>
some do full evaluation always (or have an option for that)
<rm>
not C compilers specifically
<Turl>
the standard dictates 'short circuit' right to left evaluation for C iirc
<mnemoc>
stuff like a(i++, i++) is undefined. but 1 || foo or 0 && foo are safe
<mnemoc>
in C
<Turl>
any other comment before I resubmit?
<mnemoc>
i've seen people prefering if (1 /* foo()*/) { /* NOTE... */ ...but....
<aexl>
i can confirm that focaltech touchscreen works when i disable multi-touch in ft5x_ts.h. could i commit after disabling #if 0 the line #endif and editing the /*comments*/ of course? or do you also wanna keep it for the future?
<mnemoc>
there is a related patch on the ML waiting for a v2 which I fear will never come
<zub>
Marex: I'm not Czech. Ja su z Moravy. :-P
<Marex>
zub: FM ;-)
<Marex>
rm: static int i = 2 || 1 / 0;
<aexl>
mnemoc: was that for me?
<Marex>
rm: what's the value of i ... do you know ? %^)
<Marex>
(yes, that's from C / ISO9989)
<mnemoc>
aexl: yes, there is a patch for multi-touch in ft5x_ts in abandoned on the ML
Quarx has quit []
<aexl>
abandoned? from when?
<Marex>
rm: or rather ... do you know why ? :-)
<Turl>
mnemoc: rm okay, resent
<aexl>
or what do you mean with "v2 which I fear will never come"
<mnemoc>
Marex: 2 || 1 / 0; <--- how is this evaluated? you really lost me there
<Marex>
mnemoc: %^)
<mnemoc>
aexl: a v2 is the second version of a patch
<mnemoc>
aexl: after the first session of feedback
<Marex>
mnemoc: it's actually quite easy ... what do you think ?
<Turl>
Marex: 2?
* Turl
needs to reread the operator part on K&R
<Marex>
Turl: yep
<Marex>
it's 2, indeed
<mnemoc>
i thought 2 too.... but candide (a C evaluating bot) disagrees
<mnemoc>
19:43:48 <mnemoc> cc printf("%d\n", 2 || 1/0);
<mnemoc>
zub: using the stock bootloader you'll also need to revert the previous commit
<raver2046>
i have mailed lundman about that
<wingrime>
?
aexl has quit [Ping timeout: 245 seconds]
<wingrime>
is ts glitches on power adaptor fixable?
pawel5870 has quit [Ping timeout: 246 seconds]
gimli has quit [Quit: Verlassend]
<hno>
zub, there is a wiki for my uboot with instructions.
wingrime has quit [Ping timeout: 246 seconds]
<hno>
zub, sun4i is the old tree.
wingrime has joined #arm-netbook
<hno>
wingrime, what glitches on power adapter?
<wingrime>
ts recive false touches
afgdsgh has joined #arm-netbook
<hno>
wingrime, sounds like a hardware error in your hardware. Or a bad power adapter sending lots of noise.
<mnemoc>
even my N7 has TS troubles when using it charging from time to time...
<zub>
hno: I read it. and I thought I followed it. It indeed mentions the "other" (apart form sun4i) targets, but as none was "my" target (mk802+), I assumed I need sun4i. But it seems I want mele_a1000
<rm>
from what I know all kinds of touch screens and touch pads get erratic touch behaviour while connected to lousy power adapters
<RaYmAn>
atmel even promotes their newest touchscreen tech as revolutionary that eliminates the false touches issue =P
<RaYmAn>
(which indicates it must be rather common)
<Turl>
it is
<Turl>
esp with chinese chargers :)
<jelly-home>
hell I get that with my htc phone sometimes
<hno>
zub, we don't have sufficient information about mk802+ yet.
<wingrime>
look like adapter issue but i cann't charge tablet from pc
AndChat|589056 has joined #arm-netbook
<wingrime>
not enough current?
wingrime has quit [Read error: Connection reset by peer]
<zub>
hno: 99cf18c86319cf66e088ffda30d2d27c565cab27 seems to break u-boot on my mk802+ ... not sure how, but it does
<zub>
hno: when I change CONFIG_SYS_U_BOOT_MAX_SIZE_SECTORS to the old value (1000), it works (as in, I get u-boot prompt)
<zub>
hno: with current state (200), I get: Jumping to U-Boot... and then silence and nothing
<zub>
hno: while u-boot.bin size is: 151552 bytes... and the comment says: /* 256KB, enough for a full u-boot.bin */ so it should be enough (?)
<zub>
yeah! I'm booting bleeding edge :)
<zub>
hno: I got the device, so if I could help in any way, let me know
<AndChat|589056>
where i can get docs?
<hno>
zub, pelase boot u-boot from nand and execute the following
AndChat|589056 is now known as wingrime
pawel5870 has joined #arm-netbook
<hno>
md.b 0x42400000 0x82d0
<hno>
and send the output to me.
<zub>
m'kay, I'll try
<zub>
haven't touched the nand so far
<zub>
how does the nand boot work? is it the same u-boot binary? where do I need to put it?
<mnemoc>
we don't have reliable info for mk802+ yet
<mnemoc>
zub: get boot1 doing `md.b 0x42400000 0x82d0` as hno said
<zub>
I guess I can go home now...
<wingrime>
any one have non woking ft5x ts on custom kernel
<wingrime>
?
revident has joined #arm-netbook
<hno>
zub, booting from nand means booting what is installed in nand from the factory, without sdcard.
<zub>
hno: ah, I thought I need to get some (recent/yours) version of u-boot there first
<hno>
zub, uhum... right. that commit is quite borked.
<mnemoc>
.oO
<zub>
Hit any key to stop autoboot: 0
<zub>
configured to 0, seem impossible to break... so I need to hack the env (in NAND) (?)
<hno>
that's evil of them.
<mnemoc>
the env is raw in nandb iirc
<zub>
ok
<hno>
you can drop a new u-boot binary in nanda and reboot.
<zub>
maybe I can jsut nook it and try default... as 1st attempt
<zub>
nuke
<zub>
aha
<zub>
ok
<mnemoc>
make a backup :)
<hno>
but you likely need to livesuite to recover.
<hno>
if you want to run the factory android crap again.
<zub>
hm, does not uSD have priority? so if nand is all ...ed up, can't I just boot via uSD and then write stuff into nand?
<mnemoc>
hno: btw, he is the first victim of removing allwinner boot hacks from 3.4
<hno>
yes, if you have a working system from uSD with nand support.
<zub>
I see /dev/nand? so I'd hope yes :_)
<hno>
then you are all set
<hno>
except that we don't yet have access to the area we need for a proper SPL profile..
<drachensun>
wingrime you have to disable multitouch
<drachensun>
aexl was just discussing that patch today
<drachensun>
that I abandoned :-)
<drachensun>
so aexl if you are still here feel free to take it over
rellla has joined #arm-netbook
<wingrime>
i bring ts with changes thar make it work but it not compatible to others
<wingrime>
i don't know why but 21irq is hardcoded
<mnemoc>
drachensun == npeacock ?
<wingrime>
on my hw it uses irq11
<drachensun>
yup
<drachensun>
mnemoc: yup
<hno>
zub, ok, SPL should be fixed now to actually load the full u-boot.
<wingrime>
pb11 gpio pin in fex
<drachensun>
wingrime: ok, did you need the multitouh fix too?
<wingrime>
nope
<drachensun>
and I thought it used the fex look up for that
<wingrime>
see driver
<mnemoc>
hno: should I try it on my cubie again? this morning I got trapped on an infinite crashing loop
<wingrime>
irq21 hardcoded
<wingrime>
in define
<hno>
yes, it can not have worked, only loaded first 100KB of u.-boot.bin.
<wingrime>
i used IDA to see that
<mnemoc>
hno: :D
<wingrime>
and make driver work
<wingrime>
also i add power port simply support
* hno
should do actual SD boots more often. To lazy and JTAG boots most the time.
<drachensun>
let me take a look real quick, I was interested it mixing the mulitouch if possible
<RaYmAn>
hno: did you ever get fel boots working?
<wingrime>
i have to set power port gpio to 1
<hno>
RaYmAn, I haven't tried. But should be quite doable.
<wingrime>
without this i2c bus not saw my ts
<wingrime>
looklike ts power controled with gpio "power port"
<zub>
hno: just FYI: I verified the fix, works on my mk802+
<zub>
thx
<hno>
RaYmAn, first prio is getting SPL stable, which seems to be really close now.
<wingrime>
but i have bug
<wingrime>
ts work not after every resume
<hno>
zub, thanks. Was a bit worried for a while there may be deeper issues lurking, and it's a stupid error like this..
<wingrime>
drachensun: what problems are you have with ts?
rellla has quit [Quit: rellla]
<wingrime>
8j
<mnemoc>
hno: confirmed on the cubieboard too
<drachensun>
wingrime: Mine it now working in single fine in single touch, after resumes, no problems. I would like to have multi touch support so it can be late integrated to actually use gestures.
<drachensun>
so if you had another fix to that kept multitouch I wanted to know
<drachensun>
but I just looked at my fex file and 21 seems to be what my system is trying to use
<techn>
there is patch for problem you are describing
<mnemoc>
techn: ok to merge as-is?
<techn>
drachensun: wingrime, or is there ? :)
<hno>
now I just need to implement scaling up pll1 again, and should be pretty done code wise (but plenty cleanup remaining).
wingrime has quit [Ping timeout: 248 seconds]
<mnemoc>
hno: is it still starting at <300MHz? .... it takes 3.4 5s to start printing anything on my cubie :p
<hno>
mnemoc, yes.
<mnemoc>
how much is it 30? 60? :)
<hno>
?
<mnemoc>
the initial clock
<techn>
mnemoc: dunno.. since some of the changes caused illegal dependencies to axp, or is it illegal from today? .. so it could need some patching.. gotta check :/
<hno>
I don't know. Copied the register setting verbatim from boot0 and have not decoded it.
<hno>
mnemoc: 0xa1005000
<mnemoc>
techn: will rebase the wip/sun5i branch now
<techn>
mnemoc: .. with quick check didn't axp dependency from that patch.. So it could be ok.. if drachensun/wingrime could check if it helps on their problem
<drachensun>
techn: Im not sure, I'll have to test it later and see. Could be the multitouch fix is really just happen stance that removes a bug in the multitouch code.
<techn>
mnemoc: looks that there is every commit I intented.. but haven't tested if it boots
afgdsgh has quit [Read error: Connection reset by peer]
<drachensun>
I was just about to leave and all this comes up, I am pulling a copy to test now
<hno>
At least a little faster than the initial 24MHz the CPU powers on at.
<mnemoc>
:)
<mnemoc>
hno: actually faster than the 60MHz ondemand uses by default and the 300MHz mine is most of the time
<hno>
there is a little game that needs to be played with the APB0/AHB/AXI clock divisor when increasing PLL1 and need to figure out the right criterias there to make that transition safely.
<mnemoc>
and kind of makes me wonder why is it taking that long to start talking
<Turl>
hno: look at linux, you can see how to do it there
<hno>
mnemoc, is that the time from u-boot starting kernel until kernel starts talking?
<mnemoc>
hno: doesn't the PMU take care of making the clock transition safely?
<hno>
Turl, I know.
<hno>
mnemoc, the PMU manages voltages, not clocks.
<mnemoc>
hno: from Starting kernel ... to <6>Booting Linux on physical CPU 0
<mnemoc>
5s
<hno>
increasing the core voltage is a requirement to increase clock beyond 384MHz.
<mnemoc>
384MHz sounds like a fair initial clock to me
merbzt has joined #arm-netbook
<hno>
and when increasing the core clock (pll19 one need to prepare the APB0/AHB/AXI clock divisor properly as those are derived from PLL1. If not you speed those clocks too high during the transition.
<hno>
or too low.
<mnemoc>
unless you are a a13-olinuxino-micro who wants to live at max clock all day
<mnemoc>
risky game :-\
<hno>
I don't see why that needs to live at max clock really, other than possibly that cpufreq maybe requires pmu control to accept tuning the frequency.
pawel5870 has joined #arm-netbook
<hno>
but if cpufreq requires pmu control then that should be easily fixable.
<hno>
it's ok to run the cpu at higher voltage than the clock requires. Just not the reverse.
<mnemoc>
ah, ok... i thought they were bound
<mnemoc>
like not been able to go below certain level without reducing voltage
<hno>
not as far as I know. But the manual are not exacly verbose.
<mnemoc>
Turl did rewrite all that code once....
<mnemoc>
Turl!
<hno>
cpufreq?
<mnemoc>
yes
<mnemoc>
to test higher freqs
<mnemoc>
1.2-1.5GHz
<Turl>
it has a couple of bugs though, never got around to fixing it completely
<Turl>
if you want to debug it yourself, feel free to, it might be a while until I get back to it
<mnemoc>
Turl: is it by any means worse than what we have today?
<Turl>
mnemoc: what we have today doesn't stall 'randomly' when switching freqs, this does, after a while
<Turl>
so yeah it's a bit worse
<mnemoc>
ok
<Turl>
but once it's fixed it should be better
<mnemoc>
:)
<mnemoc>
i thought the bugs were also in the current code
ssvb has joined #arm-netbook
<Turl>
I haven't experienced any at least on the current if mess :P
<mnemoc>
:)
<rz2k>
hno:
<hno>
yes?
<rz2k>
I saw some new commits to u-boot allwinner, should I test them on my mele that hangs?
<rz2k>
(we've discussed previously, that one with weird boot1 header)
<hno>
Turl, looks like your code is looping over APB divisor settings when changing frequency. Old code just sets transition divisor, PLL1, target divisor.
<hno>
rz2k, mele_a1000 spl should work nowI think. Please test. Note that make target have changed to mele_a1000 not sunxi.
<Turl>
hno: the loop is 'unrolled' on the old code
<hno>
Not really. The old code jumps from min directly to max if you ask it to.
<hno>
1. Set transition APB divisor.
<rz2k>
did you base ram settings on my script bin from nanda that I supplied or from ones available at mnemoc's script.bin archive? because my mele is actually a2000 v1.6, I dont really know, but wouldnt these settings vary per model/rev, e.g. different ram installed.
<hno>
hm. but what about transitions within the same group? Looks to me as if you then fiddle with the divisors as if it was crossing a boundary
revident has quit [Quit: Ugh]
<Turl>
possibly
<Turl>
but it shouldn't hurt right?
<hno>
well, i think you have an off by one error, with one intermediarey transition too many each time.
<hno>
which means you have a bad transition there.
<hno>
it does hurt as it means there is one transition crossing a boundary without gently twiddling the divisors.
<hno>
on each frequency change, even if new == old
<hno>
but need to study more carefully without kids in my arms.
<traeak>
hehe
<traeak>
you can type this well with kids around?
<hno>
I have our youngest (5 months) in my arms
<hno>
one handed typing
<traeak>
ahh
<traeak>
my 2yro usually is directly interfering
<traeak>
or stealing stuf and running off with it
<hno>
the baby isn't doing much yet, but we have 5 more..
<traeak>
so hopefully they are just stealking stuff from each other :-p
<hno>
the 2 year old wants to type when sitting in my lab, or vander off with various items..the 3.5 year old steals stuff from everyone and trows it around. The older onse (5 year and up) leaves my stuff salone but often gets into small fights sand come running...
<hno>
all typical behaviours for kids.
<mnemoc>
:)
<mnemoc>
and no one is ever bored :p
<traeak>
hehe
<hno>
always something going.
drachensun has joined #arm-netbook
raver2046 has quit [Ping timeout: 245 seconds]
mysteryname has joined #arm-netbook
<hno>
Aha, I don't really need to be that careful with the clocks in u-boot. No I/O running while switching frequency here.
tzafrir has quit [Ping timeout: 240 seconds]
Entropy512_ has joined #arm-netbook
merbzt has quit [Ping timeout: 248 seconds]
<mnemoc>
hno: uart? :)
eflatun has joined #arm-netbook
<lundman>
whats 1.3beta though? 1.2 was terrible
<mnemoc>
someone should make phonixcard images from Turl's CM for our meles...