<Skaag>
My load is always at least at 1.0, never below
<Skaag>
is that normal for the platform?
Soru has joined #linux-sunxi
<arokux2>
libv: is there 2D accel for sunxi? (with blobs or whatever)
\\Mr_C\\ has joined #linux-sunxi
Soru__ has quit [Ping timeout: 245 seconds]
<deasy>
Skaag, what do you use? board+distro
<deasy>
it's not normal
<Skaag>
A13 from Olimex, Debian Wheezy
<Skaag>
Kernel 3.4.61
<Skaag>
I wonder then what's causing this
arokux2 has quit [Quit: Leaving]
<apo>
actually, I seem to have that too
<apo>
hardly any disk I/O, no CPU load, but a load of 1.06...
<apo>
I'm not really bothered, but it is curious
<Skaag>
yes I'm not bothered either, everything seems to be working fine, but still it's weird that even when nothing is happening, the load is always 1.0
<Skaag>
it's too "round"
<Skaag>
iotop shows nothing serious going on
Soru__ has joined #linux-sunxi
Soru has quit [Ping timeout: 272 seconds]
egbert has quit [Disconnected by services]
egbert has joined #linux-sunxi
<deasy>
Skaag, no one processus use the cpu?
<deasy>
can you try with htop ?
<Skaag>
yep, trying
<Skaag>
while running htop, it is the topmost process always
<Skaag>
all the rest use less cpu than htop
<Skaag>
while htop itself uses 4% cpu most of the time
<deasy>
yes like me htop always on top or almost here on my x86
<Skaag>
this is not a cpu thing. htop shows cpu never crosses 20%.
<Skaag>
and it's mostly around 7% really
<deasy>
hmm
<Skaag>
this is something else, where the linux kernel thinks load is at 1.0
<Skaag>
load is a combination of things, it's not really just cpu usage
<deasy>
it's something about how much time a processus have to wait for be executed
<deasy>
i think
<deasy>
i have read something about it
<Skaag>
I thought iotop would show me what's doing it, but even iotop is showing that most of the time, there is almost no i/o happening
<Skaag>
yes exactly
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<apo>
not how much time
<apo>
just the number of processes waiting for something
<libv>
arokux: no blobs, but yes, there is
<libv>
arokux: there is 2d acceleration support in the kernel, it is not a perfect engine apparently, but i think that fbturbo makes use of it
<libv>
hrm, infobox.
pitillo has quit [Ping timeout: 244 seconds]
<apo>
22 ? D 0:37 [usb-hardware-sc]
<apo>
ah.
<apo>
Skaag: still here?
<libv>
mnemoc, oliver, Turl, arokux: any idea where infobox is defined?
pitillo has joined #linux-sunxi
eebrah_ has joined #linux-sunxi
TheSeven has quit [Disconnected by services]
[7] has joined #linux-sunxi
Gerwin_J has quit [Quit: Gerwin_J]
JohnDoe_71Rus has joined #linux-sunxi
<Skaag>
apo, here now
eebrah_ has quit [Read error: Connection reset by peer]
<Skaag>
apo, right, the usb-hardware-sc has a status of 'D' (and is the only one with that status)
<Skaag>
why is that?
<Skaag>
(D = Disk Sleep)
eebrah_ has joined #linux-sunxi
JohnDoe_71Rus has quit [Read error: Connection reset by peer]
JohnDoe_71Rus has joined #linux-sunxi
fredy has quit [Ping timeout: 245 seconds]
fredy has joined #linux-sunxi
<chkktri>
morning, in kernel config what does "CONFIG_ANDROID" enables?
<JohnDoe_71Rus>
think kernel for android
<chkktri>
guess that' not required for other distros then
<mnemoc>
libv: infobox is a template
<libv>
mnemoc: where is it defined? so i can find out all possible entries
<libv>
mnemoc: the best of luck with your upcoming interview, btw
<libv>
i will add LCD, touch and camera entries to that
<libv>
and will document this template, in yet another wiki page, where i will try to explain decent board page creation
JohnDoe_71Rus has joined #linux-sunxi
Soru__ has quit [Ping timeout: 264 seconds]
<mnemoc>
libv: usually template docu is done in the <noinclude></noinclude> part, while the template itself goes in <includeonly></includeonly>
<mnemoc>
of the same Template:
<libv>
ok
<mnemoc>
in wikipedia they go a little farther and made a template to include {{Foo/Documentation}} into [[Template:Foo]]
<mnemoc>
but inside [[Template:Foo]]'s <noinclude> block obviusly
<libv>
mnemoc: i am hoping that the new board howto will clear some things up, and will make properly structured board pages more consistent and useful
<mnemoc>
+1
<libv>
but for something like that, templates will get in the way imho
<mnemoc>
a mix of both :)
<libv>
for the infobox, it seems just what the doctor ordered :)
<mnemoc>
you don't really want to have fully template generated pages
<mnemoc>
but the infobox can do cool things, even auto-categorization based on the fields
<libv>
yeah, that will only limit things and make people less likely to create such a page
<libv>
i will create a fully empty example page though, so people can just copy/paste
<mnemoc>
:)
eebrah_ has quit [Read error: Connection reset by peer]
eebrah_ has joined #linux-sunxi
wolfy has joined #linux-sunxi
<plaes>
mkimage utility comes from u-boot?
<plaes>
the one that's required to build uImage?
<chkktri>
indeed, there's one but i guess you can find it in repositories too
<plaes>
ah.. Gentoo has u-boot-tools package
BluesBoy has joined #linux-sunxi
soul has joined #linux-sunxi
soul is now known as Guest31355
<plaes>
hrm.. the link to ubuntu-alip is not working
<oliv3r>
Turl: it did? which one? got a link so I can check? though that's the sun6i driver from rhombus, we do have the 3.4 lichee dump for sun7i which does have a gmac driver
eebrah_ has quit [Ping timeout: 264 seconds]
<oliv3r>
Skaag: old bug, was fixed ages ago if i'm not mistaken; or was it a config issue based on a script.bin settings? I don't recall; but it's on the ML/wiki
NotRelevant has joined #linux-sunxi
n01 has joined #linux-sunxi
rellla has joined #linux-sunxi
<NotRelevant>
Just for the record: On 20131110,1205 (see logs) I was trying to access my 3.5" HDD from Android (cb_a20_android42_v1.05-8188eu-en_US.img). The HDD had a GPT partition table and... it didn't work. I've tried later the same but with a MBR partiton and it worked. Also I've noticed that the "-t type" argument to the mount command is REQUIRED (the help suggests that is optional). The /etc/vold.fstab was not modified. Example: m
eebrah_ has joined #linux-sunxi
<plaes>
NotRelevant: your message was cut after "Example:"
<NotRelevant>
... The /etc/vold.fstab was not modified. Example: mount -t ext4 /dev/block/sda1 /mnt/sata
jinzo has joined #linux-sunxi
<libv>
plaes: yeah, linaro likes to move things about a lot
<plaes>
\o/
* plaes
has linaro-nano running :)
<libv>
plaes: nice, but alip might have been more useful :)
<libv>
really cool though
Guest31355 has quit [Read error: Operation timed out]
<plaes>
too bad I don't have microUSB plug so I could try out keyboard :S
<libv>
ouch, without serial or a keyboard, you probably are unable to do anything
<libv>
are you sure that no usb OTG cable was provided with your device?
<plaes>
at home :)
<libv>
ah :)
soul has joined #linux-sunxi
soul is now known as Guest63124
arete74 has joined #linux-sunxi
<chkktri>
lol, there is sun4i package in AUR, but it uses tmpfs for extracting kernel repo ...which is about 2GB huge
HeHoPMaJIeH has joined #linux-sunxi
NotRelevant has left #linux-sunxi ["."]
<plaes>
libv: linaro doesn't seem to be providing alip packages anymore :S
<plaes>
mm.. on this page: linux-sunxi.org/Bootable_SD_card
<plaes>
in the "Partitioning" section
<arokux>
chkktri: we have fixed audio, if you need it now, you need to apply a patch. do you need it now?
<plaes>
doesn't the first command kill the bootloader?
<plaes>
`dd if=/dev/zero of=$card bs=1M count=1`
<arokux>
plaes: right, linaro project deleted this image....
eebrah_ has joined #linux-sunxi
<arokux>
plaes: can you please look if there is a new image and fix our link?
notmart has joined #linux-sunxi
<arokux>
WarheadsSE: take a look at this by libv ---> " there is 2d acceleration support in the kernel, it is not a perfect engine apparently, but i think that fbturbo makes use of it "
<libv>
arokux: i am on it
<arokux>
libv: talking with WarheadsSE ?
<libv>
arokux: apparently, linaro decided to change their policy and their ALIP is now openembedded based
<arokux>
libv: what???
<libv>
so it no longer is a useful option for our first order approximation standard linux installation
<arokux>
libv: it was deleted because there is retention policy of 30 days or so, focus find it out.
<libv>
there is one left for each monthly release
<libv>
but it is no longer present since september
<arokux>
mnemoc: it would be so cool if the e-mails would include a link to gmane ...
<arokux>
libv: I was going to ask mnemoc to build our own, it is not difficult. everything is based on debian live and we just can take scripts from linaro
<libv>
now really, why doesn't it want to accept my pngs.
maz_ has quit [Ping timeout: 252 seconds]
<Turl>
libv: huh, I used apt-get on alip the other day o.O
vicenteH has quit [Ping timeout: 248 seconds]
<arokux>
Turl: you were puling arm packages from ubuntu
<arokux>
Turl: alip is really a customized ubuntu for arm afaik
AreaScout has joined #linux-sunxi
<libv>
Turl: alip is just a special mix with basic useful things installed, like a working X and a not too heavy wm
<libv>
arokux: i really rather doubt that this alexey eromenko is willing to actually donate any hardware
<arokux>
libv: why? :)
<libv>
arokux: he has made it quite clear and obvious that he is not interested to work with linux-sunxi
<arokux>
libv: why this e-mail then?
<libv>
oh, he just wants to whine
<chkktri>
arokux: great, i'll try it now
<n01>
mripard: for RTC do I have also to submit again the DT bindings doc?
<libv>
if he had any other intention, he would've come to linux-sunxi with his proposal
<focus>
arokux, libv: you can ask nicely in #linaro if someone will remake the latest linaro - they may do it (they are more likely to do it if is one of their partners), there is also a way to make it from one of their web based tools - i forget details now
<chkktri>
btw, i just ran old kernel which came with arch linux and everything works so much faster :\
<arokux>
chkktri: define everything
<mripard>
n01: yes, it's not been merged
<chkktri>
e.g.: mocp used to eat ~90% of CPU when playing 128kbps MP3 stream from net while with old kernel it's about 5%
<chkktri>
arokux: define where?
<arokux>
chkktri: I mean: describe what is slow.
<arokux>
chkktri: hm.. that needs investigation.....
<chkktri>
ah
notmart_ has joined #linux-sunxi
<chkktri>
well, mocp CPU usage, shell response time, maybe something else
<arokux>
mripard: there are some pulls! :)
<focus>
turl: i think (can't remember properly) i must thank you for suggestions a while back for changing uboot environment so that iteaduino board can be persuaded to boot from sata - got it working on Ubuntu 13.04 last night booting from sata :D
<arokux>
focus: well done! you probably have the biggest farm of allwinner boards :)
<focus>
printenv, setenv and saveenv works as described
<focus>
arokux: :D
<focus>
mk802 as well
<focus>
and some exynos
<n01>
mripard: ok. I'm also almost done with the lradc. But before submitting I was thinking to add also triggers and buffers
* plaes
just pushed Gemei G9 support for u-boot
<arokux>
focus: have you seen new arndale board?
<focus>
arokux: more important - they are *working* boards with big desktop distros - thanks all to you guys :)
<mnemoc>
arokux: it's not possible to pre-generate gmane URLs because they keep their own numbering
Black_Horseman has quit [Read error: Operation timed out]
<mnemoc>
I also mailed the gmane guy asking for renaming the group to gmane.linux.ports.arm.sunxi
<mnemoc>
like omap's
<arokux>
mnemoc: it'd be is just so cool to have a direct link to an e-mail
<mnemoc>
once we run our own mailing list and archive we could
<n01>
mnemoc: what do you use to read gmane nntp?
<focus>
arokux: nice - but the exynos quad cores are no faster than allwinner A20 - anything more than dual core, they all have same problems = DDR page turn speed is still 10MHz
<arokux>
focus: o_O
<focus>
unless they build one of my digital versions of quantum computer, they all deluding themselves over speed improvements
<Turl>
chkktri: maybe you weren't using cpufreq earlier and you're now?
<focus>
may be one of us make enough money and we all get together and make this new class of CPU chips
<focus>
too much work me thinks :(
<mripard>
n01: I think you don't need the buffers and triggers, but if you really want to work on something, I guess you can work on making the buttons work
<focus>
my fpga skills close to nil
<n01>
mripard: ok, then I'll submit the patch with only raw readings. I changed a lot from the vinifr version.
<chkktri>
Turl: well, i don't remember even using that, what's this?
<Turl>
focus: I don't particularly recall doing that, but you're welcome :p
<Turl>
chkktri: cpufreq scales the cpu frequency based on usage/demand
<focus>
turl: you are welcome
<focus>
i don't remember either
<focus>
:)
<mripard>
n01: but I really think the buttons support, even polled for now, is much more important than buffer/triggers
<Turl>
chkktri: so if nobody is using much cpu, it'll run slowly
<n01>
mripard: the only suggestion you gave him that I did't address is where to define which channel to use. If it is not in DT where I can put it? sysfs?
<n01>
also because in at91 is defined in DT
<mripard>
you don't have to define them.
<arokux>
focus: have you actually did some objective testing quadcore exynos vs dual core A20?
<mripard>
don't focus on AT91
<n01>
so I enable both and the user read one or the other by sysfs?
<mripard>
a) the maintainer was speaking nonsense
<mripard>
b) the hardware is quite different
<n01>
yeah I know
<mripard>
especially since ADC channels need to be muxed, while we don't
<mripard>
but yeah, just use whatever the guy is reading from sysfs
<mripard>
it's enough to know which channel to use
<n01>
ok. got it.
<libv>
heh, it really doesn't want to accept my images today.
<libv>
can anyone try to upload an image to the wiki?
<focus>
arokux: not precise testing - but general boot times and sluggishness is about the same as A10/A20.
<Turl>
libv: do you get that it's empty?
<libv>
yeah
<Turl>
libv: it's a server bug
<libv>
ok?
<Turl>
libv: /tmp misteriously just fills
<arokux>
focus: i could compare RK3188 and Cubietruck in near future.
<Turl>
even if it's empty
<dapsaille2>
hi o/ .. does someone know the max pixel clock for a13 ?
<arokux>
focus: i.e. Radxa
<focus>
arokux: where i would expect to see improvement is in applications where the page turn rates are low such as deep computing loops
<dapsaille2>
min pixel clock sorry :x
<arokux>
focus: so not if used as a compilation box?
<Turl>
libv: let me see if I can fix it
<oliv3r>
arokux: did you try my u-boot? does it make 2GB work for you?
<arokux>
oliv3r: not yet, but I've tried sun6i_gmac.c, ugly code drop haven't brought results, but it compiles.
<focus>
arokux: where it won't work is with applications such as servers servicing lots of different request, definitely not compiling (unless its got a lot of cache ram it can access), and general desktop responsive computing
<arokux>
oliv3r: maybe you can make more sense out of it.
<focus>
arokux: since desktop and servers are far more important, this whole arm thing is just not going to see improvement until the page turn speed is improved, cache size improved or some kind of architecture change
<libv>
Turl: thanks
<arokux>
focus: that is really discouraging. but for 2 x A20 the bottleneck could be intercommunitcation, couldn't it?
<focus>
arokux: even if they communicate by ethernet, it won't be bottleneck compared to the page turn limit of 10MHz
<Turl>
libv: try to upload now
<arokux>
Turl: what have you done? :p
<libv>
Turl: worked!
<arokux>
focus: o_O hm... I'd like to see some real tests, have you seen any?
aep has quit [Read error: Connection reset by peer]
<oliv3r>
arokux: that's the a31 gmac driver, use the one from lichee-3.4 for a20 (sun7i)
<Turl>
arokux: killed some zombieish shell scripts with deleted file handles on /tmp
<arokux>
oliv3r: that was u-boot
<focus>
arokux: what happens is that the CPUs need to do a lot of jumping and looping - and if that is into the same page, then page turn does not occur - also if big data transfers, then page turn happens only a limited number of times because DMA controller is in charge of the data transfer - so 2 x A20 a lot better communicating through ethernet than direct interprocessor communications
<oliv3r>
arokux: yeah that sun6i stuff is really old
<Turl>
libv: great
aep has joined #linux-sunxi
<focus>
arokux: so 2 x A20 servicing a web page by one sending the pages, and one sending the images is 2 x performance compared to one quad core A31
<focus>
arokux: likewise with database - if you can split the query into two, then each cpu can go access the hard disk (ssd better) and grab half the answers more quickly than one quad core or 8 core.
<arokux>
oliv3r: no GMAC in there.
<focus>
arokuk: that division of data is however very difficult - need to implement transactions - a lot harder and time consuming
<oliv3r>
arokux: idk; but not important, it's 'somewhere' and that's good enough :)
<arokux>
oliv3r: no, that is not good :(
<oliv3r>
Turl: ok first I liked the CCM_MBUS_CTRL_N_X define change, but now i'm doubting it, it's confusing :p
naobsd has joined #linux-sunxi
<Turl>
oliv3r: I'd use log2 but I dunno if it's defined on uboot :p
<oliv3r>
probably not :)
<oliv3r>
as long as it's the same on sun5i and sun7i it's good
<arokux>
focus: if quadcore are already bottlenecked, what can be said about octa core ones? :-O
<oliv3r>
well most octa aren't really octa's, they are big.LITTLE :)
eebrah_ has quit [Ping timeout: 265 seconds]
JohnDoe_71Rus has quit [Read error: Connection reset by peer]
<oliv3r>
but doesn't qualcomm I think it was say the same, there's no point in keep adding cores to improve performance
<oliv3r>
once you go beyond dualcore, the benefit becomes minimal
<plaes>
yeah, but it's a marketing push
<oliv3r>
2.2 is good enough, and 4.4 is really the maximum you should do with the number of cores
<arokux>
oliv3r: and you do not remember the reference? :p
<oliv3r>
though 2 little, 4 big is probably the sweetspot
<focus>
arokux: samsung knew this bottlneck and made it into 2 x 4 CPU sets - one set low power and other set high power - 8 core sounds impressive for marketing - but the moment they make claims it speeds up desktop by x8, the lawsuits will start flying :)
<focus>
arokux: if it really is that good, then everyone will be claiming dual cores 2x powerful and 8x cores 8x powerful - but reality is that 4x cpus are probably slower than 2x cores for general desktop computing
JohnDoe_71Rus has joined #linux-sunxi
naobsd has quit [Quit: Page closed]
<arokux>
focus: "DDR page turn speed is 10MHz", so DDR should be improved?
<libv>
what is "page turn speed" anyway?
rellla3 has joined #linux-sunxi
* arokux
has no idea
rellla3 has quit [Client Quit]
<libv>
the only reference i find to that has to do with epaper
<focus>
arokux: it can't be improved because of physics - it needs a fundamental redesign of computers (hence my suggestion digital version of quantum computer - about 5% extra silicon for 100x REAL performance improvement)
<arokux>
focus: can you plz define "page turn speed"
<focus>
arokux: DDR is organized into pages - a bit like sectors - each address location within a page is accessible in about 10ns. but once you cross to a different page then all kinds of changes need to be made to the chip to access the next page - by the time its done - its cost it 100nS or more
<focus>
arokux: so if two applications constantly bitch over which page they want - the max speed of the cpu gets limited to about 10MHz
<arokux>
focus: I see. the same technology is used on x86 platforms, right?
<libv>
focus: i really rather doubt that ddr[1-4] memory chips have any knowledge of paging
<libv>
focus: that's a higher level function implemented by the memory controller
<torbenh3>
libv: the page is a row in ram terms.
<torbenh3>
anyways, caches are there to mitigate this :>
<focus>
arokux: static ram doesn't have this problem - but they consume vast amounts of power and chip space - so the physics problem is that you can have high speed at high power or high density with low power and page turning speed rate limited
<Turl>
focus: wouldn't cpu caches solve the problem though?
<libv>
focus: bearing in mind that multi core systems tend to have cache subsystems to match?
<focus>
turl: yes and no - if 4 applications can run on quad core with 4 caches, then fine, but as soon as it turns into a database or something larger than the cache, the bitching over which page to access happens and suddenly the cpu is rate limited to 10MHz again as the total throughput rate from the cores to the DDR
<libv>
where does this 10MHz "hard" number come from?
<focus>
it it sh row access time - not a hard number and its worse than that - about 7MHz?
<libv>
focus: wouldn't one have the same issue with a single core system as well?
<focus>
libv: 100%!!
<libv>
focus: sure, you do not get 8x the performance out of an 8 core system, that's a given
<libv>
but stating that it is completely irrelevant is not true
<focus>
libv: if you could boast 6x performance out of an 8x core you still get sued - its about 5% better than dual core I would imagine :(
<focus>
and that if the applications let you
<libv>
focus: again, a hard number
<focus>
libv: approximate number
<mripard>
focus: that number also assumes that you have all the caches, prefetching, etc. disabled, and that you only do load and stores, which seems to be a pretty useless system and workload
<libv>
focus: how come it is 7MHz for every DDR module, for every layout of DDR modules, every bus, and every memory clock?
<focus>
mripard: i'm limiting myself to typical desktop applications and server application - not specialist applications that can be designed to make full use of cache and the way ddr works
popolon has quit [Quit: Quitte]
<torbenh3>
in my mind, a typical desktop application is idle 99%.
<focus>
mripard: i am assuming caches are working as expected for typical set up of desktop + server applications
<mripard>
then a typical application won't access memory at each instruction
<mripard>
and a typical application will let the CPU do its out of order execution and its prefetching to optimise all these things.
<focus>
torbenh3: i'm limiting myself to occasions when running something like database or opening a files
<torbenh3>
opeing a file ? thats waiting for the harddisk most of the time :S
<focus>
tobenh3: during file transfer the problems don't appear because DMA is in charge and random page turns are not as frequent
<focus>
torbenh3 - i meant opening something like a 10Mb html file - lots of stuff that needs to get done - or a word file etc.
<focus>
libv: the problem are down to physics
<torbenh3>
well... this is why one tries to use cache local algos.
<arokux>
DYNAMIC RAM (DRAM): CURRENT AND POSSIBLE FUTURE IMPLEMENTATIONS (April 2013)
<torbenh3>
so the conclusion is, that throwing cpus at membound problems, is not useful... oh well
<focus>
torbenh3: yep
<torbenh3>
i throw cpus at cache bound problems and, it works very well.
<arokux>
The paper contains some interesting references as: The Benefits of Quad-Core CPUs in Mobile Devices by nVidia
<torbenh3>
its very sweet to see 32core machines compiling kernels.
<focus>
arokux: they be wanting my digital version of quantum entangle memory :D - at least x100 performace for 5% silicon - i be dreaming
<focus>
torbenh3: they are better off fitting more cache - but that increases heat and energy consumption - it sudden stop being ARM and become intel :(
<Turl>
the main issue with scaling on cores is that most apps use 1 or at most 2 cores
<oliv3r>
even using 2 cores is very rare
<Turl>
so unless you're doing a highly parallel workload, scaling on freq is better
<focus>
Turl: freqency scaling works up to a point where the stuff inside cache goes faster - but soon as it involves accessing external DDR, the page turn limit limits the overal computing throughput
<arokux>
Turl: mm.. independent benchmarks would be much better.
<Turl>
focus: and if you have less cores, you can put more caches and avoid it :)
<Turl>
arokux: yeah, dunno how accurate the 170% figure is, but the concept holds
<focus>
Turl: i do all kinds of tests for years - and settle on SSD and dual core as best - anything else is wishful thinking for practical problems
<focus>
turl: cores costs practically no space - its all taken by the cache
geecko has quit [Ping timeout: 245 seconds]
<Turl>
focus: I found 4x performance increase on most of my workloads when switching from dual core to quad core/octa-thread
<Turl>
it really depends on what you're doing I suppose
<focus>
Turl: i not understand that - i have quad cores - but no faster than dual cores for most things - and yes it does depend on what is being done
<Turl>
on mobile on the other hand, not so much from going dual->quad
<Turl>
GPU, IO and memory bandwidth are the main constraints there
<focus>
some dbase things i do take half hour - not much faster with ssd on dual or quad core
<focus>
burning dvds, compressing files, moving files around, etc. etc common desktop things - hardly any difference on quad core compared to dual core
<Turl>
burning dvds is completely burner-bound
<Turl>
compressing files on parallel highly benefits from cores
<focus>
Turl: yes - burn dvd, compress files, and move files all at same time..:D
setkeh has quit [Read error: Operation timed out]
<slapin>
hi, all!
<focus>
i believe i notice improvements in compile speeds - but not sure - but that is just what I mean can't tell in real terms what benefit quad is over dual otherwise i buy quad and 8 cores tomorrow
<Turl>
focus: you risk io-starving your disk and making that dvd worthless :p or at least once upon a time that could happen
<slapin>
where is rz2k's mtd experiments?
<focus>
slapin: hi
<Turl>
hey slapin
<Turl>
slapin: did that patch I linked you the other day help with the uboot->linux mac?
<focus>
Turl: it hasn't happen yet :)
<Turl>
slapin: I did a quick test on cb2 and it seemed to be ok
<slapin>
Turl: I postponed that for a bit as I'm digging in mtd kernel stuff, but will integrate all stuff this weekend, so will look
<Turl>
focus: you can compile android in 15-20m on quad core box, vs several hours on dual core
<focus>
Turl: that is interesting to know
<Turl>
probably if I throw more ram on dual core box it'll be less drastic difference, but still perf increases
<arokux>
focus: box -- x86 or ARM?
<Turl>
slapin: ok, let me know then
setkeh has joined #linux-sunxi
<focus>
Turl: i believe latest android compile may well need 32GB RAM for this performance - true or false?
<Turl>
focus: I haven't tried yet
<Turl>
but 4.3 is usually fine with ~16G
<focus>
Turl: thanks - boss wants this looking into at some stage
<arokux>
slapin: that is one Xeon or a dual CPU set up?
<slapin>
arokux: dual Xeon on 4-way motherboard, minimal configuration I could get (for commercial project)
<slapin>
arokux: some supermicro platform
<slapin>
arokux: can't remember the details, but something low-end
<arokux>
slapin: doesn't matter, thanks for the info!
<slapin>
arokux: the xeon is good for speedy memory speedy disc controllers, and well-done parallelism, which is the best trio for build servers
<slapin>
arokux: and faster raw performance
tzafrir has quit [Ping timeout: 245 seconds]
<oliv3r>
Turl: they use 2 'octal buffer/line driver' per color channel to convert the digital RGB to analog RGB signals, very cool how they did that, so you shoudl treat it as a LCD interface i'd expect
enrico_ has joined #linux-sunxi
<slapin>
Turl: octal buffer/line driver is just resistor-based DAC, as I see
<slapin>
Turl: should be cheap and work well
<slapin>
Turl: but prone to be variadic on temperature changes
* slapin
tries to be funny
* arokux
never heard of SuperMicro before
* wens
thinks about getting a cubietruck
* dapsaille2
banghead against crt tv
<oliv3r>
SuperMicro makes awesome server boards
<oliv3r>
wens: olimex is sweeter :)
pitillo has quit [Changing host]
pitillo has joined #linux-sunxi
<dapsaille2>
supermicro offer really great motherboards
tzafrir has joined #linux-sunxi
<oliv3r>
and i think there's a few even supporting coreboot
<wens>
oliv3r: i already have a cb2. want something with gbe
* wens
thinks gmac on cb2 won't work
<oliv3r>
if you get a pin-compatible module .. it probabyl won't :p
<oliv3r>
wens: butt yeah ct is only GBE module for now
<wens>
oliv3r: i think A20 gmac pinout only supports rgmii
deasy has joined #linux-sunxi
<oliv3r>
wens: dno't think so, but the PHY used on the CB3 is the limitted rgmii only one
<wens>
but i'll keep testing
tzafrir_laptop has joined #linux-sunxi
tzafrir has quit [Ping timeout: 272 seconds]
<wens>
would be nice to have a working sample
<arokux>
wens: what will you test?
<wens>
arokux: since gmac and stmmac are almost identical, trying to get stmmac to work with gmac on sunxi-next
<arokux>
torbenh3: did some work on it too, last weekend. you both may want not to duplicate yourselves.
<wens>
arokux: were you talking to me?
<arokux>
yes (remove ":")
<arokux>
wens: nobody has started to add gmac to u-boot though. as I said, I've just dropped the code and it did not work. I'll try your suggestion and fix base address
<oliv3r>
arokux: i assume you only put in the gmac driver, not the entire u-boot from a31 right? :)
<arokux>
oliv3r: right.
<oliv3r>
good good :)
<torbenh3>
wens: yeah. using stmmac is probably the way to go.
<wens>
arokux: you'll also have to change the pinmux values, from 0x2 to 0x5, and probably add a few missing ones
<arokux>
wens: oh...
tomee^ has joined #linux-sunxi
<tomee^>
hi
<torbenh3>
if (of_device_is_compatible(np, "st,spear600-gmac") ||
<wens>
there's this wierd piece of code in gmac: "#ifdef SUN7i_GMAC_FPGA"
<wens>
wonder what it does.. the pin function mentioned in that section doesn't exist in sun7i manual
<tomee^>
does anyone here have had experience with cubietruck/cubieboard v3? it seems to be very similar, but somehow I run into various trouble when following cb2/a20 tutorials
<wens>
torbenh3: i have stmmac compiled, and dts updated. i don't know if the driver has actually initialized the gmac.
<tomee^>
very similar to cb2 that is
<wens>
mdio probing yields no phy
<torbenh3>
wens: sure, you got the pinctrl right ?
<wens>
torbenh3: looks right, PA01 ~ PA16 to gmac
<oliv3r>
wens: that's for FPGA development, when the chip was just an FPGA and not a real chip yet
<wens>
oliv3r: I see
<oliv3r>
wens: you can concider it #if 0
<chkktri>
ah, great, everything works
<wens>
torbenh3: come to think of it, maybe gmac isn't defined in pinctrl yet...
notmart has joined #linux-sunxi
<wens>
ok, just checked, it is
<arokux>
chkktri: you mean sound? what about perf?
<chkktri>
both
<arokux>
tomee^: you need to be more clear. what fails exactly, what error messages you get etc. etc. etc.
<arokux>
chkktri: how did you improve perf?
<chkktri>
CPU was indeed running at 60MHz most of time
<tomee^>
arokux: for one thing, I have doubts about the bootloader. the pre-built sdcard distros don't seem to boot (script.bin differences I guess). the "lubuntu" example distro (from cubieboard.org) does boot but is completely bare - will take ages to figure out right branches of libhybris etc.
<chkktri>
now i have a question: anybody has docs on A10's CODEC? I wonder what playback rates are supported without resampling
<arokux>
chkktri: scan the wiki man page, Hardware
<tomee^>
arokux: moreover, in your wiki, there is a lengthy tutorial involving boot.scr, recompiling the u-boot each time you modify a parameter, whereas the bootloader that comes with the "example" distro only has... 3 files. uEnv.txt, script.bin (I know about bin2fex/fex2bin) and uImage. and it works.
<arokux>
tomee^: can you show that page please?
<tomee^>
arokux: I also spent a couple hours compiling the 3.4.61+ kernel (sunxi-mali tree->cubieboard fork). it does work, but when I try to send any packet through sunxi_gmac or bcmdhd, the kernel oopses. sata and usb work fine, though... and they are also gpio-dependent I guess.
hipboi has joined #linux-sunxi
<arokux>
tomee^: cubietruck just went on sale. I do not know what comes preloaded with it, but you probably cannot expect all the other stuff to work out of the box. no network for sure. your description is also way to general for us to start improving things. I assume you are just a user, so I'm not blaming you. however if you want to get help you'd need to be specific.
<arokux>
tomee^: where have you got this kernel from?
<tomee^>
arokux: I am just a user since I just got the cubietruck, probably first batch at all (which I didn't realize when I ordered the gizmo)
<tomee^>
arokux: but I have experience with programming atmega chips so I am not a computer idiot
<arokux>
tomee^: yes, congrats, you'd need to hack a bit to get it working 100% :)
<arokux>
tomee^: we here work on the community kernel aka sunxi-3.4 and mainlining. that one you show is not supported by us, although we take the code from there to make our sunxi-3.4 even better.
<tomee^>
arokux: I understand, cubie forks your stuff, messes around, some things work, some don't, you don't support their branch, that's obvious.
<arokux>
tomee^: maybe you'll have more luck with this.
<tomee^>
arokux: but does your "vanilla" branch support cb3/ct1 or "not yet determined" ?
<arokux>
tomee^: right. but we are very interested in taking their changes and putting them into sunx-3.4.
<arokux>
tomee^: I do not know if somebody tried sunx-3.4 with ct, oliv3r ?
<tomee^>
arokux: and about that wiki page: http://linux-sunxi.org/Optimizing_system_performance#Increasing_the_memory_clock_speed, "ecause the memory clock has to be configured at the beginning of the boot process, it is not straightforward to change the memory clock speed. It involves recompiling u-boot-sunxi and re-flashing the bootloader on your SD-card."
<tomee^>
arokux: from what I learnt, it involves changing 1 parameter in script.fex which sits on an ext2 partition...
<plaes>
hmm.. wpa_supplicant cannot set essid/keys for rtl8192cu card :S
<arokux>
plaes: which board? which error?
<arokux>
tomee^: hm.. wiki might be true. which parameter can you change?
<plaes>
Gemei G9 and latest linaro-alip build
<plaes>
ioctl[SIOCSIWAP]: Operation not permitted
<plaes>
ioctl[SIOCSIWESSID]: Operation not permitted
<plaes>
etc..
<plaes>
arokux: ^^
<tomee^>
arokux: [dram_para]/dram_clk in script.fex. or is it just an INDICATOR not a SETTABLE? (I thought the purpose of script.fex was SETTING parameters...)
<arokux>
tomee^: yes, the purpose is to set params, you are correct.
<arokux>
plaes: maybe you want to become root?
<plaes>
I am
<Turl>
tomee^: you cannot normally change the dram freq using script.fex
<Turl>
it's there in case you use livesuit or so, which do read it and hardcode the bootloader to use that one
<tomee^>
Turl: then what is that line doing in my script.fex (vendor-provided) ?
<tomee^>
Turl: so most of things are settable via script.fex, but some need to be hard-coded into the bootloader and those parameters are then just indicators for a compiler to copy their value when burning ?
<arokux>
moofree: Processor Model: BOXCHIP All Volunteers A20 :D
<Turl>
tomee^: yes
<Turl>
tomee^: changing the ram frequency wipes the ram contents
<Turl>
tomee^: so it can only be normally set really early on boot
<Turl>
and there's not much space to work there for a script.bin or such parser, mmc driver, fat/ext driver, etc
<tomee^>
arokux: ok, summing up: 1) I involuntarily got myself a real "hot of the press" board instead of a typical documented barebone 2) you recommend I start with your kernel tree, cubie/sunxi-3.4 branch. 3) which distribution (or prebuilt image, help me God) do you suggest? aim=working 720p video
cajg has quit [Ping timeout: 272 seconds]
<tomee^>
Turl: hmm, bad news. and will it work with bootloader on sd, or only on nand (I guess it always goes through the nand one first, right?)
<Turl>
tomee^: 1) dunno what you mean 2) yeah 3) any I suppose, linaro ubuntu or debian may be the easiest as libv has packages for those
<Turl>
there's some cubie distributions and stuff you can find and just burn to a card
JohnDoe_71Rus has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org]
<Turl>
tomee^: if you have a suitable sd, nand is never touched
<tomee^>
Turl: I have tried one of those "cubie" distros but they're for cb2... so it still involves hacking around the bootloader with unknown outcome...
<tomee^>
Turl: ad. 1) I meant I didn't realize that I am ordering a completely new board when I clicked "buy" ;)
<Turl>
tomee^: what hw do you have? I somehow thought you said cb2
<tomee^>
I thought ct1/cb3 was around for like half a year.
<Turl>
ah, ct
<Turl>
more like a month or two if anything
<tomee^>
but it seems that production really started on ~20-th oct or something
<tomee^>
torbenh3: thx, I somehow skipped fedora-related stuff so far (debian/ubuntu user mostly).
<tomee^>
torbenh3: do you know what's the status of cedarx/cedarv/libve/hybris in this distro?
<tomee^>
torbenh3: preinstalled, available prebuilt, or need to recompile everything?
<torbenh3>
tomee^: nope. i dont know the root password of it, and i didnt bother yet. after looking at the kernel source, i found it offensive, and i need to get mainline ready :>
<tomee^>
so I'm still in the dark ;)
<tomee^>
I tried the "let's build it from the grounds up" approach since I have access to lots of CPU
<tomee^>
but since the custom-crafted kernel crashes on network all plans collapsed...
<tomee^>
"The Wifi driver haven't been merged to linux-sunxi kernel, so the module doesn't work currently on r3 version. But a working wifi driver can be found at github.com/cubieboard2/linux-sunxi branch sunxi-3.4-ct-v101. "
<tomee^>
another great news ;)
Black_Horseman has joined #linux-sunxi
Black_Horseman has joined #linux-sunxi
Black_Horseman has quit [Changing host]
<tomee^>
ok, one last question for now. your libhybris wiki page is not very idiot-proof. should I copy all of android's NAND system/lib to /system/lib, to the hybris build dir or where? I tried mkdir -p /system/lib && cp -aR /nandd/lib /system/lib && make install, but the test binaries crash ;(
eebrah_ has quit [Read error: Connection reset by peer]
<dapsaille2>
i will look at source code .. if i'm not back in one hour please tell yo my wife that i loved her :)
<deasy>
French guy detected
<dapsaille2>
damn .... spotted ..
<deasy>
salut :D
<dapsaille2>
salut :)
<deasy>
you should be registered on the wiki and complete it as it seems hard ton configure uboot
<deasy>
to*
<dapsaille2>
ok but .. it may not be a -sync -vsync problem as i can read in xorg.log that the modeline is -hsync +vsync
<dapsaille2>
-vsync sorry ..
<deasy>
you use vga out?
<dapsaille2>
yes, on an a13
<dapsaille2>
dumb question . which uboot fb olinuxino a13 use ?
<dapsaille2>
forget it please .. i'm looking in board dir
boycottg00gle has joined #linux-sunxi
<slapin>
hi, all!
<dapsaille2>
o/
<slapin>
do we have support for graphics in u-boot?
<oliv3r>
slapin: no
vicenteH has quit [Ping timeout: 240 seconds]
<oliv3r>
slapin: wingrime was playing with it, but couldn't get it to work (yet)
<dapsaille2>
well .. i cannot find which fb is used for a13 in u-boot for tweak hsync vsync
<oliv3r>
but it's def. a 'nice to have' feature
<oliv3r>
dapsaille2: there is no graphics/fb support in u-boot
<oliv3r>
until the framebuffer driver kicks in, your blind
<dapsaille2>
ok .. so i need to tweak linux-sunxi sources for modify hsync vsync polarity ?
<oliv3r>
dapsaille2: if it can't be done with ffex, then yes
<dapsaille2>
thanks
<dapsaille2>
looks like fex don't provide such option .
<dapsaille2>
humm .. i cannot find anuything related to hsync and vsync in lcd drivers .. but in hdmi drivers it exist .. sounds bad for me
<Skaag>
oliv3r, that's good news that it's been fixed. do you know how to search for it in the ML? I would love to find it.
<Skaag>
I mean some keyword to find it by
<oliv3r>
i think on the wiki i'll be easier to find
<oliv3r>
gotta restart screen; brb
oliv3r has left #linux-sunxi [#linux-sunxi]
Fusing has joined #linux-sunxi
<mnemoc>
gmane has search
nove has joined #linux-sunxi
oliv3r has joined #linux-sunxi
<oliv3r>
there, 8 bit color goodness, must beter then 3 bit colors!
tomee^ has quit [Ping timeout: 264 seconds]
jemk has joined #linux-sunxi
tomee^ has joined #linux-sunxi
vicenteH has joined #linux-sunxi
setkeh has quit [Ping timeout: 252 seconds]
setkeh has joined #linux-sunxi
hipboi has quit [Remote host closed the connection]
fredy has quit [Excess Flood]
fredy has joined #linux-sunxi
<arokux>
tomee^: still there? our kernel is sunxi-3.4, you'd like to grab stage/sunxi-3.4 and tell us what is missing there if comparing to the manufactures kernel.
<tomee^>
I am aware of that, but haven't tried out yet
<wolfy>
would the use of fbturbo be of any help if I need to display on a cubieboard2 several mjpeg video feeds ( security cameras ) in the same time ?
<arokux>
libv: ^
Gerwin_J has quit [Quit: Gerwin_J]
<libv>
tomee^: kernel and u-boot are cross
<libv>
tomee^: other bits you usually do not need to bother with
<libv>
fbturbo has ubuntu packages, then there is the mali binary drivers for which there is a howto, then there is the cedarx stuff...
<libv>
but for an armhf enabled distribution, the amount of work needed is usually limited
<libv>
i personally compile libump/fbturbo/mesa on the target (as i have a quad core exynos to do it for me)
<tomee^>
i see
HeHoPMaJIeH has quit [Quit: Leaving]
<torbenh3>
i use an arm chroot with qemu-user-static if i need to build big things.
<mnemoc>
<3 qemu-user-static
<libv>
how fast would qemu be?
<libv>
i have a quad A9 running at 2GHz with 2GB ram, and, sadly, a hd connected over usb2.0
paulk-collins has joined #linux-sunxi
<mnemoc>
libv: much faster than doing the same directly on target hardware :p
<libv>
my quad core 2nd gen intel core at 2.5GHz, emulated, probably is not worth the hassle
<libv>
mesa packaged in 21 minutes?
<torbenh3>
libv: exynos is going to be faster, i guess. but io with qemu-user is native...
<libv>
yeah, io is where exynos is really let down, only getting like 23-25MBps
<libv>
but that usually only matters when running make install or creating the actual packages
<torbenh3>
how big is mesa, compared to qt ?
<libv>
on an A10, qemu definitely wins, on A20 one can start to wonder whether having 2 different installations (one qemu, one on target) really makes sense for the few bits that you need to compile, on exynos44, i really do not think it is worth the hassle
<libv>
torbenh3: with debian packages built and 2-3 copies of the binaries in the tree, 2.3G
<torbenh3>
well... the buildsystem is falling out of our image generation process ;) so i have them lying around for free.
montjoie[home] has joined #linux-sunxi
<deasy>
arokux, which distro you use for build kernel?
<tomee^>
libv: plus, 16 SSDs in RAID10 are a bit faster than one SATA drive, or even SSD drive, if I was crazy enough to spend 200$ for hdd to plug into a $100 board
<tomee^>
libv: and this makes A LOT of difference in stuff like git clone, ./configure etc.
Gerwin_J has joined #linux-sunxi
<arokux>
tomee^: git clone is network bound, though?
<tomee^>
arokux: git clone, right, ok, but syncs, merging, all other file operations are not
eebrah_ has quit [Ping timeout: 244 seconds]
tzafrir_laptop has quit [Read error: Operation timed out]
fredy has quit [Excess Flood]
Gerwin_J has quit [Quit: Gerwin_J]
fredy has joined #linux-sunxi
<WarheadsSE>
tomee^: distcc
<tomee^>
distcc on what?
<WarheadsSE>
that should solve your host vs dest problem
<tomee^>
on 1 cubieboard? :-)
<WarheadsSE>
do the preparse/configure ont he device, offload the CC to an x86 machine
<arokux2>
mnemoc: can you throw that query at DB and see what happens?
<n01>
mripard: since we are using iio just for raw reading does it make sense to put the sampling rate in sysfs?
<n01>
I'd leave it in dts
<oliv3r>
wouldn't it be nice to change the rate on the fly?
<oliv3r>
in standby, sample very rarly
<arokux2>
oliv3r: will you u-boot 2GiB-patch be accepted upstream?
<n01>
oliv3r: it's pretty useless IMO since you are doing raw readings
<Upgreydd>
mnemoc: so i should made bootable sd card and extract arch into rootfs... and it should working? Should i build uboot, script.bin and kernel or just use this from arch linux?
<arokux2>
your*
<n01>
cat /sys/blabla/adc
<oliv3r>
arokux2: i don't see why not; and it's not even all 100% needed; but there's a bug, and I asked 3 times about it, they kept arguing about random stuff
<arokux2>
oliv3r: :)
<oliv3r>
so hno needs weigh in his heavier voice
<arokux2>
oliv3r: you do the same some time ;)
eebrah_ has quit [Read error: Connection reset by peer]
<arokux2>
hno: has abandoned us
<Upgreydd>
arokux2: so i should made bootable sd card and extract arch into rootfs... and it should working? Should i build uboot, script.bin and kernel or just use this from arch linux?
<n01>
oliv3r: I mean that the adc is started only when you read from sysfs
<oliv3r>
if you read the mailinglist, they haven't even looked at the patch, just some bla bla
<oliv3r>
n01: for lradc? but how will it trigger an interrupt while asleep to wake the device ;)
<arokux2>
oliv3r: you are right, you at least look at the patch
<n01>
oliv3r: atm I'm not intersted in triggers
<Upgreydd>
deasy: so i should made bootable sd card and extract arch into rootfs... and it should working? Should i build uboot, script.bin and kernel or just use this from arch linux?
<oliv3r>
arokux2: i try to :)
<ssvb>
dapsaille2: hotswitching the screen resolution is possible even now
<dapsaille2>
hot switching ?
<dapsaille2>
in fb or xorg ?
eebrah_ has joined #linux-sunxi
<dapsaille2>
didn't find anything related to this, i'm trying to use a 15.6khz monitor trough vga connector and i only get a black screen .. so if i could use Xorg modelines instead of editing script.fex everytime it would be perfect for me :)
<ssvb>
dapsaille2: if you are looking for a perfect solution, then there is none yet :)
<dapsaille2>
hehe :)
<arokux2>
Upgreydd: you build by your self: script.bin, u-boot, kernel. then you combine it with rootfs from Arch world.
<dapsaille2>
anyway, if i can change resolution without rebooting, it's ok
<dapsaille2>
and .. did you tried to use an old crt tv trough vga ? i cannot find anyone who have tried this so ... i'm not sure it possible and i lack knowledge of arm/sunxi world ...
<dapsaille2>
of course, i can do it with x86 debian ^^
<Upgreydd>
arokux2: but i should use script.fex from sunxi board or somehow take it from my device? :P
<arokux2>
Upgreydd: no if you build you kernel by yourself
<Upgreydd>
arokux2: ok, thank you :]
<ssvb>
dapsaille2: for vga I only used fex so far
<arokux2>
Upgreydd: nice, so somebody has done that bit for you
<mnemoc>
arokux2: it's stock REL1_20 + extensions, all running from git/svn and updating automatically
<dapsaille2>
damn .. no hdmi output on my a13 olinuxino ... did you used a 15khz monitor trough vga with fex parameters ?
<arokux2>
mnemoc: now even categorytree renders incorrectly i.e. different
<arokux2>
mnemoc: not + and trees
<arokux2>
no*
<mnemoc>
upstream updates I suppose
drachensun has joined #linux-sunxi
<dapsaille2>
anyway hdmi to rgb scart/jama is not possible i think ^^
<arokux2>
mnemoc: disregard, it was browser cache!
<mnemoc>
:D
<arokux2>
jemk: have you tried GMAC with sunxi-3.4?
<ssvb>
dapsaille2: what is the complete modeline for that 15khz mode? I used the standard resolutions which are configurable via fex (800x600, 1280x720 and 1920x1080), and they work
<dapsaille2>
there is many .. final goal is to emulate old video systems ... exemple SNK Neo Geo = Modeline "384x224" 7.552440 384 400 435 486 224 240 243 259 -hsync -vsync
<hramrach>
arokux2: gmac wfm with sunxi 3.4
Upgreydd has quit [Ping timeout: 250 seconds]
<dapsaille2>
i've successfully transmlated to .fex parameters (confirmer by xorg.logs) but i only get a black screen ... i don't know where i fail :x
<arokux2>
hramrach: post patches please, so it can be added
<hramrach>
I did not make them
<hramrach>
they are in cubieboard repo
<arokux2>
hramrach: maybe you have github branch for that?
<arokux2>
hramrach: so you just grabbed that kernel?
<hramrach>
github.com/cubieboard/linux-sunxi
<hramrach>
Branch cubie/sunxi-3.4
<hramrach>
no, just the gmac patch
<hramrach>
tried the wifi patch as well but that one did not work
<mnemoc>
wth..... preempt_count() = 0; <--- is this thing legal C????
<hramrach>
actually it may be a macro which is then technically legal
<hramrach>
possibly
* arokux2
thinks the same
<mnemoc>
if the function is returning a pointer, it's "legal".... but still wtf-ish
<hramrach>
no, you would need an asterisk then
<mnemoc>
true
<hramrach>
if it compiles it's legal ;-)
<arokux2>
mnemoc: will it compile?
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<kz1>
I'm trying to use /proc/sys/kernel/random/uuid for the value in /sys/class/android_usb/android0/iSerial "on init". I have added a script but the call to service doesn't seem to kick in or it's overwritten. is it possible to set a dynamic variable in init.rc files?
<hramrach>
mnemoc: but while you are looking at patches try cherrypicking the gmac patch from cubieboard repo
geecko has joined #linux-sunxi
FR^2 has quit [Quit: Connection reset by peer]
<arokux2>
hramrach: how was ct so far for you?
<mnemoc>
hramrach: I prefer to merge it from a pre-cherrypicked branch for-amery on your repo ;-)
<arokux2>
mnemoc: what are you doing yourself? :p
<mnemoc>
CC drivers/power/axp_power/axp-mfd.o
<mnemoc>
LD drivers/power/axp_power/built-in.o
<mnemoc>
LD drivers/power/built-in.o
<mnemoc>
not even a warning.... uh
Fusing has joined #linux-sunxi
<mnemoc>
arokux2: script.bin to platform_device generation
<hramrach>
if they patch it in already can't they send the patches too?
guizamboni has joined #linux-sunxi
<arokux2>
hramrach: I am asking myself the same
<arokux2>
hramrach: but they donated boards to us.. so not that bad.
<mnemoc>
even olimex fails to do that
pacopad has quit [Quit: pacopad]
<arokux2>
even??
<arokux2>
why even?
<mnemoc>
cubietech is chinese
<mnemoc>
olimex is an oshw european company
<mnemoc>
and they just keep their own tree, patch and fix things and post on their blog
<arokux2>
so what? :)
<arokux2>
yeah I do not like their blogging culture too.. they probably want to attract users to their web site
<mnemoc>
if a european vendor doesn't understand the value of submiting the f* patches, don't expect the chinese ones to do so
<mnemoc>
fixes that solve problems on olinuxino boards are already for people owning olinuxinos.... no need to "attract" them to their web site... they are already customers!
dapsaille2 has quit [Ping timeout: 250 seconds]
<arokux2>
mnemoc: hm.. cannot disagree with you
<oliv3r>
heh, i always small uSD
<mripard>
n01: no, it's not hardware description, and every user might want a different one, or change it on the fly (just for example to make shannon happy)
deasy has joined #linux-sunxi
<oliv3r>
i thought olimex sends those very limitted of number of patches they generate to devs
<mnemoc>
we are giving olimex boards a prime location in our wiki's Main Page, but they still prefered to create their own *closed* wiki
<mnemoc>
and document there
<oliv3r>
but they aren't software people, they dn't write patches really
<mnemoc>
they do... and publish the fixes on their github and blog
* arokux2
never saw mnemoc so angry
<mnemoc>
ugly as hell fixes, but fixes anyway
<mnemoc>
arokux2: not angry. frustrated
<arokux2>
<wens> arokux: you'll also have to change the pinmux values, from 0x2 to 0x5, and probably add a few missing ones
<arokux2>
what could that mean?
<mnemoc>
i'm not a business person and that's probably why I fail to see their PoV
<mnemoc>
arokux2: pinmux value is the "feature" selected for each pin