BJfreeman has quit [Read error: Connection reset by peer]
_BJFreeman is now known as BJfreeman
egbert has quit [Disconnected by services]
egbert has joined #linux-sunxi
ganbold_ is now known as ganbold
ZetaNeta has quit [Ping timeout: 260 seconds]
ZetaNeta has joined #linux-sunxi
ZetaNeta is now known as Guest51602
resistencio has joined #linux-sunxi
<resistencio>
HI i downloaded the lastest sunxi kernel from hitub but i'm unable to find option module needed for 3G usb modem
<resistencio>
isn't where is supoused to be
<resistencio>
drivers usb serial
resistencio has quit [Quit: Ex-Chat]
eebrah_ has joined #linux-sunxi
[7] has quit [Disconnected by services]
TheSeven has joined #linux-sunxi
BJfreeman has quit [Quit: had a good time]
eebrah_ has quit [Ping timeout: 252 seconds]
wowon_ has quit [Quit: Leaving]
eebrah_ has joined #linux-sunxi
eebrah_ has quit [Ping timeout: 248 seconds]
TheSeven has quit [*.net *.split]
theOzzieRat has quit [*.net *.split]
forcev_ has quit [*.net *.split]
awafaa has quit [*.net *.split]
ssvb has quit [*.net *.split]
bsdfox has quit [*.net *.split]
MadSpark has quit [*.net *.split]
TheSeven has joined #linux-sunxi
forcev_ has joined #linux-sunxi
theOzzieRat has joined #linux-sunxi
awafaa has joined #linux-sunxi
MadSpark has joined #linux-sunxi
ssvb has joined #linux-sunxi
bsdfox has joined #linux-sunxi
eebrah_ has joined #linux-sunxi
shineworld has joined #linux-sunxi
eebrah_ has quit [Ping timeout: 256 seconds]
rellla has joined #linux-sunxi
<oliv3r>
resistencio maybe your missing a dependacy? maybe your 3G modem doesn't have ARM drivers?
<jelly-home>
maybe whoever built the kernel forgot to define CONFIG_USB_SERIAL_OPTION
<jelly-home>
there's more than one kernel build out there with a really meager choice of drivers built
<oliv3r>
yeah but the default config for sunxi is pretty reasonable
<oliv3r>
also, 'make linux-config' is the next possible step I suppose :)
<oliv3r>
*back to documenting on the wiki :S*
rz2k has joined #linux-sunxi
* rellla
wonders if oliv3r dreamed about the wiki ...
<oliv3r>
heh, i hate editing the wiki :S
<oliv3r>
so it's nightmares if anything
n01 has joined #linux-sunxi
<oliv3r>
n01: !
<oliv3r>
n01 md 0x01c20c90 1 says that WDT_KEY = 0; so is the default really 'something? or does it need to be turned on first?
<n01>
'morning oliv3r :)
<oliv3r>
also, good morning :p
<n01>
TMR_WDT_CTRL is supposed to be 0
<n01>
"Restarts the watchdog timer" so if you write 1 the wdt is simply restarted
<n01>
but maybe I did not understood your question
<n01>
*understand
<oliv3r>
WDT_KEY
<oliv3r>
TMR_WDT_KEY 1:12 Read/Write 0x0a57
<oliv3r>
that one :)
<n01>
ooooohh
<oliv3r>
i do a md in u-boot, but it reads 0x00
<oliv3r>
is it write only?
<n01>
yep AFAIK the key is used only when you reset the wdt (write)
<oliv3r>
(assuming the key is actually needed to arm the timer)
<oliv3r>
ok so 'reading' will always read as 0, writing it is required to arm the timer
<oliv3r>
or, disarm rather (reset)
<oliv3r>
so you'd do writel(reg, WDT_RESET | WDT_KEY);
<n01>
yes, even though mripard said that the key could be not really used on A10
<n01>
WDT_KEY is shifted by 1
e-ndy has quit [Remote host closed the connection]
<oliv3r>
yeah, should be relativly easy to test then, specially with more hardware to test this on (a20!)
<n01>
yup
<oliv3r>
writel(reg, WDT_RESET | WDT_KEY << 1);
<n01>
I hope in the giveaway :D
mturquette has quit [Quit: leaving]
<oliv3r>
:p
<n01>
yes but: (1) check for the precedence between | and << and (2) since you don't know what is in bits 13:31 it is better to read the reg content value before modifying it
<oliv3r>
it was an example :p
eebrah|away is now known as eebrah
<oliv3r>
i agree, a weak one :p
<oliv3r>
it was basically a nerdy way to say, so you supply the key simultaniously to restarting the WDT or it won't work.
<oliv3r>
and now you made me type it all out anyway :p
<n01>
:D
<oliv3r>
there, tried to wiki-fy it
<oliv3r>
if you could double check it for obvious mistakes, that'd be swell
<n01>
no prob
<oliv3r>
then again, i made a typo, sec
<oliv3r>
there
eebrah_ has joined #linux-sunxi
tavishb has joined #linux-sunxi
e-ndy has joined #linux-sunxi
hipboi has joined #linux-sunxi
rz2k has quit [Read error: Connection reset by peer]
eebrah_ has quit [Ping timeout: 264 seconds]
<oliv3r>
n01: did you manage to get me a working kernel? :)
<n01>
ups ... you have to remind me after 7pm when I go home
_BJFreeman has joined #linux-sunxi
<oliv3r>
heh
<oliv3r>
i was busy writing wiki's anyway till 12
<oliv3r>
:p
<oliv3r>
just thought of it now
_BJFreeman is now known as BJfreeman
<n01>
oliv3r: anyway AFAIR it should be enough to use multi_v7_defconfig, enabling sun4i-a10-cubieboard.dts and eventually to use initramfs
<oliv3r>
yeah, but i get no kernel output :( so it may run, but i don't see
<oliv3r>
i used setenv bootargs earlyprintk console=ttyS0,115200
<oliv3r>
hno: those strings are at the very least interesting
<oliv3r>
'ar100 notify ac327 wake up event detect', ac327 is the internal name for a31?
<n01>
mripard: can I leave the names I used?
<oliv3r>
or for the CPU core or something?
<mripard>
n01: yeah, definitely
<n01>
good
<oliv3r>
well if better names are possible ... :p
<n01>
I'll name one register OLIVER_REG
<oliv3r>
:p
<hno>
oliv3r, ac327 seems to be the code name of the ARM part of the SoC. A31 is the marketing name for the complete SoC. aw1633 seems to be the code name of the SoC.
<oliv3r>
ah, i see
<hno>
A10 is AW1623, A13 AW1625
<oliv3r>
we don't know about A20 yet do we?
<hno>
aw165x I think.
<hno>
config ARCH_SUN6I
<hno>
bool "Allwinner Ltd. AW163x family"
<hno>
bool "Allwinner Ltd. AW165x family"
<hno>
config ARCH_SUN7I
<oliv3r>
ah yes
<mnemoc>
hno: can this little endian openrisc be worked out with standard tools?
<oliv3r>
mripard: yay! i can get output, though fails at booting
<Turl>
oliv3r: btw, I've pushed sunxi-next to linux-sunxi and verified it to boot
<oliv3r>
sunxi-next is the 3.9 branch; right?
<oliv3r>
well right now, my u-boot refuses to work :( so I think my kernel works, but uboot is broken :(
<n01>
oliv3r: do you use cubieboard_defconfig on sunxi-current?
<oliv3r>
i buildusing the BSP, so I do assume so yes
<n01>
why? just compile from sources
<oliv3r>
UBOOT_CONFIG=cubieboard
<oliv3r>
the BSP 'just compiles from sources'
<oliv3r>
the whole 'eat your own dog food' thing
<n01>
why don't you compile by yourself? :)
<mnemoc>
you can edit the chosen_board.mk file manually if you like
<oliv3r>
n01: because that would be stupid :p
<oliv3r>
the bsp doesn't do much else, cd u-boot; make
<oliv3r>
it just sets up some variables
<n01>
not at all ... at least you would have the full control over compilation IMHO
<oliv3r>
but what would I wanna control?
<oliv3r>
the BSP is just some fluff to ease things
<mnemoc>
saves a lot of typing
<mnemoc>
but you still in contorl
<n01>
ok, neverming
<n01>
*d
<oliv3r>
anyway, since we recommend it sort of to users, we gotta test it too :)
<oliv3r>
n01: have you ever looked at the bsp?
<n01>
nope
<oliv3r>
u-boot, linux-sunxi are git-submodules
<Turl>
oliv3r: 3.10-rcN+ actually
<oliv3r>
the 'script' simple is a Makefile
<oliv3r>
it's nothing speciall really :)
<oliv3r>
not some secret sauce that patches things :)
<oliv3r>
n01: one really big advantage is, it seperates the generated output files from the source, so they don't get mixed. anyway, this is what the BSP uses to build u-boot:
<oliv3r>
make -C u-boot-sunxi all O=/silo/build/sunxi-bsp/build/cubieboard-u-boot CROSS_COMPILE=arm-pc-linux-gnueabi- -j8
<oliv3r>
n01: so I don't really see any reason to replace 'make u-boot' with 'cd u-boot-sunxi; make .....' :p
<mnemoc>
the O= is very useful to keep your source tree clean
<oliv3r>
ok I lie, when rm -r the build dir; i get:
<oliv3r>
make -C u-boot-sunxi cubieboard_config O=/silo/build/sunxi-bsp/build/cubieboard-u-boot CROSS_COMPILE=arm-pc-linux-gnueabi- -j8
<Turl>
what toolchain is that?
<mnemoc>
gentoo?
<oliv3r>
yes
<oliv3r>
gcc version 4.6.3 (Gentoo 4.6.3 p1.11, pie-0.5.2)
<oliv3r>
well you define the name of the toolchain yourself :p
<n01>
better using codesourcery
<n01>
I had _a lot_ of problems with (i.e.) linaro
<oliv3r>
nah, this is managed as a package by gentoo
<oliv3r>
haven't had any issue YET with the toolchain
<oliv3r>
it just works
<oliv3r>
i hear a lot of talk about linaro
<tavishb>
what exactly does 3:two_same_screen_diff_contets<screen0,screen1,fb0> (for A10) mean? two outputs are enabled with same contents(1 framebuffer)?
<oliv3r>
gcc version 4.6.3 (Gentoo 4.6.3 p1.11, pie-0.5.2)
<oliv3r>
erm
<oliv3r>
Mentor Graphics decided to make a change in their site. Now in order to download their toolchain you have to register an account, which means giving them first name, last name, email, phone, company, job function, and country. So, even though the download is gratis, they are not delivering the tools as freely as before.
<oliv3r>
that's allready a reason not to use codesourcery
<mnemoc>
we have a direct link to download codesourcery's toolchain in the wiki
<oliv3r>
but are you 'legally and officially' allowed to obtain it that way?
<oliv3r>
anyhow, noty, not intersted
<mnemoc>
sure, it's gpl
<mnemoc>
the registration is just a annoynance, not a requirement
<mnemoc>
they keep the files in a public server anyway
<oliv3r>
but if there's gcc-arm; why would I want to use their toolchain? I mean, if their patches are so good; they should land pustream eventually?
<mnemoc>
only because it's the toolchain normally used by manufacturers
<mnemoc>
and if you want to inject things it's better to match their toolchain
hipboi has quit [Quit: Leaving]
<oliv3r>
that i can understand, but i have no reason to do anything like that
<oliv3r>
i'll do some git bisecting tonight if I have time
<oliv3r>
building, puting on www, downloading, is a little tiresome
<mnemoc>
neither do I. I was only adding completeness about mentor's annoyance page to get the codesourcery toolchain
<oliv3r>
lol thanks :D
<mnemoc>
:p
<oliv3r>
still, i odn' thave a working u-boot :( sunxi doesn't work with dtb; sunxi-current doesn't run period :S
<oliv3r>
and was planning on documenting/testing more stuff today!
tinti has joined #linux-sunxi
<mnemoc>
uhm
<hno>
oliv3r, sunxi-current worked for me on cubieboard yesterday. But I'll try again.
rz2k has joined #linux-sunxi
<oliv3r>
hno: it could be the DD job, i have to run it at work here, from XP, under cygwin
<oliv3r>
so it may fail there
<oliv3r>
but it displays the header, initial ram etc
<oliv3r>
i'll run some comparisions to initial values that i found with our SPL, and compare it against that Allwinner u-boot +boot0 +boot1; something to do :)
<hno>
the source is in my github repo. But tHere is only one completely unrelated change compared to linux-sunxi repo.
shineworld has joined #linux-sunxi
<oliv3r>
hno: i'll try that, again could be that DD didn't work properly
<oliv3r>
hno: is it worth, comparing md/mw/mtest results from allwinner stock nand boot, vs our SPL booted from mmc?
<hno>
comparing what?
<hno>
what are you looking for?
<rz2k>
hno: is it possible to establish a data channel thru FEL and flash mtdblock partitions? I checked yuq's work and he uses tftp for getting images. for sun5i tftp is not accesible.
<rz2k>
since you got u-boot working with FEL, I believe we can do it
<hno>
rz2k, I not only have u-boot working but a full boot of Linux.
<oliv3r>
rz2k: i know he does some magic using livesuit, basically to bypass the randomizer
LoCoZeNoz_ZUE has left #linux-sunxi ["Too loud for my irssi"]
<oliv3r>
hno: when you worked with slapin on mtd, you tried to get it to work without DMA, right? Does yuq's also do this? or he only implemented it with dma, and if we'd want it without, we're still equally stuck?
<oliv3r>
anyhow, this fel boot is amazingly awesome to quickly test different kernels/u-boots etc; makes development a lot nicer
<hno>
It's faster to tftp than fel, but sure.
<oliv3r>
you mean because fel writes take longer?
<oliv3r>
well my tablet for example, doesn't have ethernet :p
<oliv3r>
also, i don' thave to 'copy binaries to tftp root, reboot etc'
<oliv3r>
sure that can all be scripted
<oliv3r>
but this just seems easier/better :)
<oliv3r>
even if the actual write over USB is slower then over ethernet (which isn't available always everywhere)
<oliv3r>
also, with tftp, you need console access (also lacking on my tablet) to initiate tftp
<oliv3r>
just need usb console somewhere mixed in :D
<hno>
kernel can have console on an usb gadged I think. But we have no usb gadged driver for u-boot yet.
<oliv3r>
I might have a microsd breakoutboard before that happens though
<mripard>
looks like more sunxi-based boards are coming :)
<hno>
oliv3r, took me 19 seconds to load fel-boot + u-boot + script.bin + kernel + initrd (total 5.7MB) over FEL.
<oliv3r>
mripard: $0!
<oliv3r>
hno: ok that's quite slow :( but still workable i suppose
<oliv3r>
mripard: looks like it'll be A20; with gigabit for the 'mediabox'
<hno>
Well, it's probably faster than switching SD cards anyway.
<mripard>
oliv3r: I think we can play the SoC guess game
<mripard>
oliv3r: yep, maybe
<oliv3r>
also 2gb ram
<mripard>
with an A10 for the basic, and an A31 for the professionnal
<oliv3r>
Mali400MP2 GPU (Dual Core)
<oliv3r>
does also smell A20
<hno>
1.2G Dual Core ARM cortex-A8 processor ???
<oliv3r>
probably copy paste fail
<oliv3r>
cortex-A7 i'd guess
<oliv3r>
the 'profesionall' fixes this
<oliv3r>
also smells of A31 as mripard said
<oliv3r>
but 35 for the torrentbox version
<mripard>
the torrentbox could be an A13, given that it's also the cheapest
<oliv3r>
that's raspberry pi pricein
<mripard>
but the network speed doesn't match
<oliv3r>
A13 compares quite well to the pi i gues
<mripard>
and the SATA
<oliv3r>
maybe it's a10 anyway?
<libv>
bullboard doesn't inspire much confidence
<oliv3r>
480Mhz memory, i'm curious if they have awesome routing, or just did a simple test/estimate
<libv>
15k for designing 4 different boards?
<libv>
and then he buggers up the A20 arm core type
<Turl>
oliv3r: can't most devices run at 480?
<oliv3r>
i suppose 3 boards are nearly identical, all a10/a20
<libv>
no, not really
<oliv3r>
Turl: I thought 480 was tricky sometimes, i thought 360 is 'most' some do 40x, cubieboard does 480, and could do 50x i belive
<libv>
torrentbox is an a13
<oliv3r>
libv: with sata and gigabit?
<libv>
heh, vga output...
<oliv3r>
usefull for your KVM in your serverroom :p
<libv>
a13 doesn't have vga
<oliv3r>
so A10 more likly?
<oliv3r>
i still think there's some copy pasting errors there
<Turl>
why would you do a torrent box without ethernet?
<libv>
also looks like he is not privy to the a10 not really clocking to 1.2GHz
<oliv3r>
900 Mhz CPU + 300 Mhz GPU!
<libv>
:)
<libv>
bullboard is full of it.
<hno>
libv, a13 can drive VGA with a little "LCD" output glue. See olimex board.
<oliv3r>
there, made a comment to clarify things
<libv>
oliv3r: urgh, now you helped this guy along in trapping some stupid indiegogoers
<libv>
hno: to some extent
<libv>
i had a dog of a time bringing that up 2 weeks ago
<oliv3r>
hey, maybe it's legit!
<oliv3r>
i wouldn't be supprised if its rebranded boards
<oliv3r>
cubieboard with different branding :p
<libv>
oliv3r: not for those prices
<oliv3r>
download cubieboard specs, change logo; send to fab?
<libv>
again, this a hungarian, and he's claiming that he is desiging _and_ shipping 4 different boards for 15k
<libv>
cubieboi had a board design before he did indiegogo
<libv>
and his revenue has been an order of magnitude bigger so far
<libv>
all on a single board
<libv>
with experience and connections in china
<libv>
bullboard is full of it.
<libv>
i just hope nobody halfway clued treads in it
<oliv3r>
well it's a private message for ow, so i'm curious to his answers initially
<oliv3r>
we can burn the witch after :p
<ssvb>
DidiBsAs: hi, did you ping me the other day?
<DidiBsAs>
Hey ssvb, yes I did. Got your nick from Youtube. I want to improve video performance on a mele a2000. And you suggested to increase memory speed to 480mhz I think. I'm trying to do that building my own uboot now.
<n01>
ssvb: are you a youtube star? didn't know :)
<DidiBsAs>
oliv3r, I'm using ssh and the NAND OS is loading instead of the SD OS. Don't know if there's some kind of boot log I can see.
<oliv3r>
you need to connect uart in that case to see why it's not booting the SD
<n01>
DidiBsAs: have you uart to see u-boot log?
<oliv3r>
might be it simply doesn't like the SD :)
<ssvb>
DidiBsAs: if you already have some image that boots properly from the SD card, you can just dd sunxi-spl.bin and u-boot.bin to the right location on it
<DidiBsAs>
Nope, don't have uart.
<ssvb>
DidiBsAs: I don't have much experience with NAND myself, starting with just SD card should be much easier
<DidiBsAs>
ssvb, sounds like a good idea, will try that.
<DidiBsAs>
I tried 800x600 1024x768 and 1366x768 and it looked pretty much the same, but I know more pixels = more processing required. Still, the youtube demo you posted looked flawless.
<ssvb>
you mentioned running some office applications, they may be slow and CPU hungry themselves
<DidiBsAs>
yes, but considering the performance I have now, I don't think too much improvement is needed, it looks like a video problem mostly
eebrah is now known as eebrah|away
tinti has quit [Read error: Connection reset by peer]
<ssvb>
DidiBsAs: that's easy to verify, you can profile your system and check what exactly is using most of the CPU
<ssvb>
DidiBsAs: if it's Xorg process, then the graphics is likely the bottleneck
<DidiBsAs>
Right, using top I saw that moving windows around a little bit takes Xorg to use almost all % of CPU.
<DidiBsAs>
I read a lot that there are not open source drivers for Mali, what is sunxi using then?
<ssvb>
Mali accelerates 3D graphics, the things like rotating cubes or first person shooters
<lunra>
afaik copies are still not accelerated by the X11 driver (and probably never will be)
<lunra>
You can reduce the problem of that by configuring your WM to only draw the outlines
<ssvb>
lunra: copies can be accelerated using G2D, but this needs a reasonably recent kernel
<lunra>
and disable your CPU meter, it's more trouble than it's worth unless you're a developer who needs to know whether his program is blocking for disk or in a busy loop in a pinch ;)
<ssvb>
DidiBsAs: when you are moving your windows, the CPU is maxed, but the overall performance should be reasonably fine
<ssvb>
DidiBsAs: but when you are working with the application (to type text or do something else with it), the bottleneck contributing to slowness may be in the application itself
<DidiBsAs>
Yes, it is possible.
<ssvb>
DidiBsAs: also to make sure that moving the windows is not pegging the CPU, you may try to enable G2D acceleration
<DidiBsAs>
How do I enable G2D acceleration?
<ssvb>
Option "AccelMethod" "G2D"
<DidiBsAs>
Right, on xorg config.
<ssvb>
in the "Device" section of xorg.conf
ganbold_ has joined #linux-sunxi
<ssvb>
but if you have an old kernel, moving windows will cause screen corruption
<DidiBsAs>
G2D acceleration is being done by CPU, not Mali?
<ssvb>
no, it's a dedicated 2D accelerator
<ssvb>
part of Allwinner A10 SoC
<DidiBsAs>
I have so much to learn...
<ssvb>
Mali is a dedicated 3D accelerator
<DidiBsAs>
I know that, but I thought there was no 2d accelerator.
<ssvb>
and CedarX is a dedicated video decoding/encoding accelerator
<ssvb>
in the case if somebody wants to watch 1080p movies
<DidiBsAs>
Looks like the SD image I had is not using g2d, do you think screen corruption is the reason why it's not there?
<ssvb>
you got the screen corruption?
<DidiBsAs>
not yet
<ssvb>
ok, just try it first and then ask :)
<ssvb>
the chances are you have a recent kernel already
<ssvb>
DidiBsAs: which office application exactly are you using? I may try to profile it
<DidiBsAs>
libreoffice
<DidiBsAs>
calc and writer
<DidiBsAs>
you mean version?
<ssvb>
well, knowing the version also would not harm
<DidiBsAs>
ssvb, LibreOffice 3.6
<DidiBsAs>
Pardon my ignorance, just cloned a10-meminfo, how do I run it?
<ssvb>
"gcc -static -o a10-meminfo-static a10-meminfo.c" and then "./a10-meminfo-static"
<ssvb>
as explained in the comments inside of a10-meminfo.c
<DidiBsAs>
:$
<ssvb>
hno: DidiBsAs has Mele A2000 which is already well supported, I only suggested to use a10-meminfo to verify the actual memory clock speed
<mnemoc>
script.bin usually has several NULL fields, which we need to fill (using a10meminfo)
vinifm has joined #linux-sunxi
<slapin_nb>
hi, all!
<slapin_nb>
do anybody have insight on meaning of DDR3 settings of A10/A13?
<hno>
yes.
<slapin_nb>
hno: I need to know what I can fine-tune there
<ssvb>
DidiBsAs: ignore this wiki page, it is for the people who have some weird new Android device based on Allwinner A10 and want to add support for it to u-boot
<slapin_nb>
hno: and which options do I have re bus width and per chip bus width
<DidiBsAs>
got it, ssvb
<slapin_nb>
hno: and is per chip delay is implemented within controller
e-ndy has quit [Remote host closed the connection]
<hno>
bus width is 8/16/32 (32 on A10 only).
<hno>
it does have some kind of built in timing training, but know othing about the timing tuning registers, those are only given in binary form in script.bin with a reference that you get these from allwinner.
<slapin_nb>
hno: can I ask allwinner to get the data for me, or should I sign NDA with them?
shineworld has quit [Quit: Leaving]
<slapin_nb>
hno: I want to use 1 x 256M x 16 chip in big version, 2 x 128 x 8 in small version and 1 x 128M x 8 in minimal version, so I want to know what should I change in timing configurations
* slapin_nb
wonders if he can enable getty in android like on OLinuXino on a stick to gather info and not bother with GUI
<hno>
Most boards I have seen is using the same timing.
<hno>
only different I/O width.
<ssvb>
DidiBsAs: ok, will try to build libreoffice 3.6, it may take a while :)
<DidiBsAs>
You are awesome, man. Thanks.
<ssvb>
but my hunch is that it might be something like a spellchecker or whatever else causing the slowness
<ssvb>
keep in mind that Allwinner A10 is not particularly fast in terms of processing power, maybe something like a 15 years old Celeron running at ~600MHz
<ssvb>
if changing the screen resolution does not help much, it's very likely the graphics is not the bottleneck
e-ndy has joined #linux-sunxi
<ssvb>
DidiBsAs: also it might make sense to chance cpufreq governor to "performance" or tweak "ondemand" as described at http://linux-sunxi.org/Cpufreq
<ssvb>
unless you have tried this already
<DidiBsAs>
nop, haven't yet. I need to write down all the things I need to try.
<luoyi>
anyone have interest on writing a libva wrapper with cedarx ?
<luoyi>
I think maybe this is the right way the make most video&audio app can use the power of cedarx
<ssvb>
luoyi: let me guess, you? :)
<luoyi>
studying libva src now. and still don't have idea on it
<luoyi>
ssvb: you are cross-compiling the libre ?
<luoyi>
I think it may cost few hours to get the job done
<ssvb>
luoyi: natively compiling on a dualcore Cortex-A15, expecting it to take at least half a day
<luoyi>
ArchLinuxArm already have the libreoffice package ...
<luoyi>
ssvb: which distro are you work with ?
<ssvb>
I prefer a build with all the symbols and debugging information
<ssvb>
gentoo
<luoyi>
so cool ... haha
<luoyi>
and do you think the symbols and debug info will hurt the performance ?
hramrach_ has joined #linux-sunxi
<ssvb>
no, that's just additional extra information which does not affect performance but greatly simplifies profiling and debugging
<luoyi>
dualcore A15, you mean Samsung's Exynos 5250 ?
<ssvb>
yes
<luoyi>
chromebook
<luoyi>
and the A15 binary code can work correctly on the AllWinner A10 ?
<lunra>
I compiled linux-sunxi stage-3.0 on my chromebook, took less than 2 hours
<ssvb>
it's the other way around, I'm compiling for -march=cortex-a8 and it works correctly on the chromebook
<ssvb>
-mcpu=cortex-8
<lunra>
then again I just realised libreoffice is like, libreoffice sized.
<lunra>
the A15 has integer division which iirc the a8 does not
<ssvb>
yeah, libreoffice is a huge monster, people must have a lot of faith in A10 to even try libreoffice on it :)
<DidiBsAs>
ssvb, so you've gave it a try? is it a no-go then?
<ssvb>
I gave it a try on my desktop computer, it's not particularly lightweight
<ssvb>
on ARM it will take hours to get it compiled, that's currently in progress
<DidiBsAs>
rz2k, scripts from cnx-softweare.com guide are creating the hardware packs, there's something wrong there probably, will try manually now
<rz2k>
scripts from cnx software are outdated
<ssvb>
the third-party blogs/articles tend to be outdated or incorrect, it's a good idea to keep all the information in linux-sunxi wiki and make sure that it is right
<ssvb>
just try to dd sunxi-spl.bin first, this may be enough
<ssvb>
but then you also may want a recent kernel, and we don't know what you have installed at the moment...
<ssvb>
so sunxi-bsp might be the best option in the end
<DidiBsAs>
I need quantal develop respositories for gcc-4.7-arm-linux-gnueabihf-base, g++-4.7-arm-linux-gnueabihf, etc right?
<oliv3r>
if you don't have a cross compiler installed allready
<oliv3r>
if you are building on debian/gentoo; you should be able to install a arm cross compiler easily
n01 has joined #linux-sunxi
<oliv3r>
hno: u-boot with a real dd works fine; cygwin's dd doesn't seem to write u-boot properly
<n01>
oliv3r: did u solve?
<oliv3r>
yeah, dd from a real os
<oliv3r>
and now installed terminal on the stock nand
<oliv3r>
so can use that too
<oliv3r>
didn't have networking at work
<oliv3r>
and no terminal means useless android
<n01>
what from google I/O?
<oliv3r>
huh?
<oliv3r>
my cubieboard has android in stock nand
<oliv3r>
but no terminal.apk
<oliv3r>
otherwise i would have dded from there
<n01>
lol .... I mean any news from the conference :D
<oliv3r>
is it today?
<mnemoc>
started ~2h ago
<n01>
yep
<oliv3r>
and no news leaks yet?
<mnemoc>
nothing about new android versions
<oliv3r>
so what DO they talk about
<oliv3r>
all we care is new android stuff or new google hardware
<oliv3r>
if they don't share either
<oliv3r>
who cares? :p
<mnemoc>
they'll sell unlocked nexus-ized samsung s4... and gave away chromebook pixels
<n01>
uhm ... chromebook could be interesting
<n01>
still a bit too expensive
<mnemoc>
the pixel is intel-powered and way overpriced to be interesting
<mnemoc>
i was waiting for a $500-600 variant with exynos5410 inside
<oliv3r>
yeah but the entry fee is like 1500 USD? so how is that 'giving away pixels'
<oliv3r>
or a $300 AMD one :p
<mnemoc>
oliv3r: they gave one to each attendee
<mnemoc>
the rest is mostly only about google services and cross platform interaction
* mnemoc
wanted a 10h pixel with exynos5 octa :'(
<oliv3r>
well the entry fee is like 1500 USD, so you are buying a pixel mostly :p
<mnemoc>
:p
<mnemoc>
so the event becomes free
<oliv3r>
if you have that kind of cash laying around
<n01>
oliv3r: the company usually pays
<mnemoc>
not mine...
<n01>
mine neither O_O
<mnemoc>
but yes, companies pay the entry fee and individuals get the gifts
<mnemoc>
for those lucky enough to work for such companies
<oliv3r>
yeah :(
<oliv3r>
linux-sunxi should start collecting donations :D
<mnemoc>
to go to google io? :p
<mnemoc>
kind of unrelated...
<oliv3r>
our kernels run android!
<n01>
??
<n01>
mine not for sure :)
<mnemoc>
+can
<oliv3r>
matter of fact, isn't it still the 'android kernel'?
<oliv3r>
so, good reason!
<DidiBsAs>
u-boot-sunxi.git is supposed to be cloned inside sunxi-bsp, right?
<mnemoc>
make u-boot-sunxi/.git should do the cloning
<mnemoc>
or as part of the `make u-boot` chain
rz2k has quit [Read error: Connection reset by peer]
theOzzieRat has quit [Ping timeout: 264 seconds]
theOzzieRat has joined #linux-sunxi
<Yakuz>
btw. was it solved what caused the high traffic on the wiki ?
<mnemoc>
not yet
<mnemoc>
it's around 130GB per day
<Yakuz>
just recalled it was an issue a week ago or so
<mnemoc>
tried to attack it using a reverse proxy in front to control the abuse of connections, but the result was awful performance for everyone
<DidiBsAs>
I suck at compiling u-boot, tried 3 different methods already.
<mnemoc>
so now I'm using the spare time to move everything to the new server, cancel the old, and then enable extra (>5TB/M) bandwidth at full speed
lkcl has quit [Ping timeout: 245 seconds]
LoCoZeNoz_ZUE has joined #linux-sunxi
<LoCoZeNoz_ZUE>
Hello. Does anyone know where to buy/get a replacement touch screen for iNet-97F? Becouse mine is partly non-functional
<LoCoZeNoz_ZUE>
Very annoying part, Top 2.2 centimeter of the screen are "display only" and when i press on em, touch is recognised at complete top making it not so annoying
<LoCoZeNoz_ZUE>
Anyone know a replacement touch for Mpman MID74C? (iNet-97F Rev2 board)
<hno>
Anyone in here with an A31 device?
lkcl has joined #linux-sunxi
<oliv3r>
hey luke
<oliv3r>
hno: drachensun has some
<ssvb>
DidiBsAs: which gcc version are you using? Also if I understand it correctly, you compile u-boot successfully, write it to the SD card but the device does not boot?
<oliv3r>
right, kernel finally, FINALLY boots :D it hangs at 'calibrating delay loop... '
<oliv3r>
though linaro's gcc is known to cause trouble
<oliv3r>
and latest gcc (forgot version) produces buggy results
<hno>
4.8
<ssvb>
DidiBsAs: have you tried sunxi-spl.bin instead of u-boot-spl.bin?
<oliv3r>
Turl: mripard what could cause a division by zero in the kernel in 3.9?
<mnemoc>
the panic should tell you
<mnemoc>
function+offset
<mnemoc>
then gdb can turn that into source + line
<DidiBsAs>
ssvb, haven't tried sunxi-spl.bin, will try in a few minutes
<oliv3r>
erm
<oliv3r>
[ 0.000000] [<c001412c>] (unwind_backtrace+0x0/0xf8) from [<c01116e4>] (Ldiv0_64+0x8/0x18)
<oliv3r>
so function is unwind_backtrace?
<techn_>
no
<oliv3r>
[ 0.000000] WARNING: at /silo/build/sunxi-bsp/linux-sunxi/kernel/time/clockevents.c:46 clockevent_delta2ns+0x84/0x90()
<oliv3r>
there then, is that what causes the booboo?
<techn_>
[<c0055c54>] (clockevents_config+0x28/0x7c) from [<c0055cbc>] (clockevents_config_and_register+0x14/0x20)
<techn_>
next in call stack is division
<techn_>
and then core dump (unwind_backtrace)
<oliv3r>
so clockevents_config causes the ldiv0
<oliv3r>
and ldiv0 calls the unwind bracktrace (the 'dump'?
<oliv3r>
clockevents_config sounds like turl's handywork :p
<techn_>
ldiv0 causes exception and exeption handler is unwind_backtrace .. not sure.. but thats how I think it
<techn_>
:)
<oliv3r>
yeah sorta yeah :)
<oliv3r>
ok so, i want to pull latest changes into my own tree, but have my last 2 commits 'on top', e.g. a rebase, but don't want ugly 'merge' commits
<oliv3r>
git pull --rebase?
<oliv3r>
oh i think i was halfway a rebase allready :)
<ssvb>
DidiBsAs: unless you have accidentally killed something else on the SD card, I think the system should boot fine
<ssvb>
DidiBsAs: I understand that it's hard to debug this stuff without serial console
<DidiBsAs>
it is, but looks like I'm closer, and I learnt a lot in the process
<ssvb>
DidiBsAs: in the worst case, revert to some SD card image that is known to boot, and then just dd your compiled spl/sunxi-spl.bin and u-boot.bin exactly as explained in the instructions from the linux-sunxi wiki
<oliv3r>
DidiBsAs: if you have a spare, 4 GiB SD card; on my tablet hansg's fedora imge worked perfectly
<DidiBsAs>
ssvb, actually, because I'm afraid I might have broke something in the process, I'm dding now a working image, to do that. Crossing fingers.
<oliv3r>
since we have the same tablets, to get you started, that should work
<oliv3r>
if that doesn't, either your SD card isn't recognized/liked by the tablet, or something else
<oliv3r>
but hansg's images should work :) in theory
<DidiBsAs>
oliv3r, when you say the same tablet, you mean we both have A10? Because Mele is not a tablet.
<oliv3r>
DidiBsAs: i thought you had a broken tablet you wanted to use; i mistaken you!
<oliv3r>
but the mele is also very well supported
<DidiBsAs>
np :)
<ssvb>
DidiBsAs: just in case, you don't need to erase and re-create partitions, just dd these two files
<oliv3r>
the really big advantage with hansg's stuff is, it's known to work; so you take that as your basis
<oliv3r>
then, if you want to swap out spl; dd it, if you want to swap u-boot; dd it; if you want to swap kernel, just swap it
<ssvb>
hramrach_: hello, this looks like a generally problematic file, it is also played as a slideshow in the native android player
paulk-desktop has quit [Quit: Ex-Chat]
<hramrach_>
but it does not crash
<hramrach_>
yes, it is problematic
<hramrach_>
I suspect it needs too much bandwidth
<ssvb>
it does not crash for me in xbmc though, do you have a backtrace?
<hramrach_>
it crashes only with hybris
<ssvb>
not for me, I get just a slideshow but no crashes
<hramrach_>
hmm, does not crash this time
n01 has quit [Read error: Connection reset by peer]
<hramrach_>
so sometimes crashes :s
n01 has joined #linux-sunxi
<ssvb>
on the other hand, after rebooting to android and now back to linux, I get blue colors on video playback in xbmc
<ssvb>
the colors are fine in vlc
<ssvb>
I remember some people reported something like this in the mailing lists, but I have not seen this before until now
<hramrach_>
I get sometimes horizontal lines in vlc
<hramrach_>
never seen the blue colors in xbmc
<drachensun>
hno: I've got A31 devices, whats up?
<drachensun>
I've got Linux booting on one too, just a minimal buildroot form though
<ssvb>
hramrach_: lol, removing this weird seventh file from the xbmc video directory fixed the blue colors issue
<hramrach_>
I get the impression that the hd playback is a lie
<hramrach_>
you can play back HD stills and nearly-stills
Yaku has joined #linux-sunxi
<hramrach_>
anthing with much change becomes a slideshow
<ssvb>
hramrach_: maybe cedarx state gets messed up when generating thumbnails and encountering this file
<hramrach_>
yes, cedar resets are not something well done in xbmc
Yakuz has quit [Ping timeout: 246 seconds]
<hramrach_>
maybe the init of that crashy video was wrong but it still worked when it was first video to get played
<hramrach_>
so there is something that carries over from previous video
<ssvb>
yeah, this seems reproducible
<ssvb>
hramrach_: can you try to put this problematic bb2.mp4 and Sintel trailer files into the same directory visible to xbmc, then open it, hover the cursor over thumbnails of these videos and then try to play the Sintel trailer?
<ssvb>
in my case the snow (and everything else) becomes blue
<hramrach_>
hmm, bb2 plays in software on i386
<hramrach_>
so with moderate amount of computing power it's decodable
<ssvb>
however mplayer complains a lot about "Missing reference picture" at least for me when playing bb2
<ssvb>
it might be not strictly standard conforming
<hramrach_>
cannot reproduce the blue snow
<ssvb>
ok, and I can't reproduce crashes
<hramrach_>
can't reproduce them either anymore
<ssvb>
was it a fresh start of xbmc when you tried to test wrong colors?
<ssvb>
maybe even rebooting would be better
von_fritz has quit [Quit: vonfritz leaves, don't panic]
<hramrach_>
no, it was running already
LoCoZeNoz_ZUE has quit [Quit: leaving]
<ssvb>
please try it again after rebooting, that's how it was reproduced for me (and xbmc may probably have some sort of thumbnails caching logic)
<hno>
drachensun, stekern in #openrisc would like to probe the openrisc core a bit to know more of what openrisc implementation is embedded in the A31.
<ssvb>
DidiBsAs: also you may first check if the system can boot with 480MHz memory clock (I guess at least booting should be always fine) and see if it improves performance
<hramrach_>
cannot reproduce
<hramrach_>
made a new directory for the test, rebooted, no blue tint
<ssvb>
thanks, I'll try to see if there is something special needed to trigger this bug
<hramrach_>
also re-compressing bb2 with x264 quiets the missing frame messages but creates artifacts
<hramrach_>
the second scene gets jumpu and is smooth in the original encode
<ssvb>
"mediainfo bb2.mp4" says that it was encoded on UTC 2004-06-13 01:10:04