<libv>
this whole repo creation thing seems relatively straightforward
<libv>
so i am quite amazed that things only went that far and not further
ssvb has joined #linux-sunxi
naobsd has quit [Quit: Page closed]
TheSeven has quit [Ping timeout: 246 seconds]
<Turl>
libv: you have an ssh account with web access, don't you? :)
<libv>
ssvb: what operatings systems are you using atm?
<Turl>
feel free to put up a repo there for testing if you wish
<libv>
Turl: not to my knowledge
<Turl>
libv: then poke mnemoc for one :)
<libv>
:)
<Turl>
libv: you need to send mnemoc your ssh pubkey, he'll do the rest
<libv>
ok
TheSeven has joined #linux-sunxi
<akaizen>
how can one easily flash a system.img to an A20 / A10 board?
<akaizen>
Do I *have* to use livesuit?
<steev>
yes
<steev>
at least if you want "easily"
Gerwin_J has quit [Quit: Gerwin_J]
<lunra>
There's a Linux livesuit (x86_64 only? iirc)... I couldn't personally get it to work though
<lunra>
(if the problem is having to start up Winblows)
Gerwin_J has joined #linux-sunxi
<akaizen>
i have a few a20 sdks
<akaizen>
im im trying to see if they actually work, the boot.img I can flash by copying it to /sdcard/ and cat boot.img > /dev/block/boot
<akaizen>
cant do the same for system since its mounted and running... i was hoping someone had a similar easy solution
<akaizen>
and i dont know how to get my A20 board into FEL boot mode ... it only has a reset button and adb reboot bootloader just does a normal reboot
<akaizen>
i can reboot to recovery but adb push system.img /dev/block/system => target disk ran out of space
kaspter has joined #linux-sunxi
speakman_ has quit [Ping timeout: 264 seconds]
speakman has joined #linux-sunxi
akaizen has quit [Remote host closed the connection]
akaizen has joined #linux-sunxi
<steev>
how does one find out what board they have in a tablet? like i have a tablet that's uh, apparently called something like a MID970 ?
<steev>
i know it's an a10
<libv>
pry it open :)
<Turl>
akaizen: push somewhere else and then cat ... > /dev/block/system
<akaizen>
Turl: I did that and it never reboots and i have to go to recovery...
<Turl>
you're probably copying a broken filesystem then
<Turl>
make sure it's not sparse ext4
<akaizen>
its the system.img thats generated by the SDK
<akaizen>
do i need to modify it somehow before pushing it?
<Turl>
yeah that's probably sparse
<Turl>
run simg2img on it first
<akaizen>
holy crap if this is hte problem i might cry
<akaizen>
is this normal? computed crc32 of 0x66f33ea9, expected 0x00000000
<Turl>
dunno, but 0 sounds like nobody computed it in the first place
speakman has quit [Ping timeout: 268 seconds]
<akaizen>
heh i should really look at more source code instead of copy+pasta
speakman has joined #linux-sunxi
<akaizen>
i've been binary diffing android builds and sys FS for like hte past week trying to figure out why these sdk images were not working ...
ykchavan has joined #linux-sunxi
<Turl>
:)
<Turl>
akaizen: btw, you're doing it from recovery right?
<akaizen>
when i boot into recovery i cant get a shell
<akaizen>
so im doing it regularly i think it should still work... im going to do a sanity check real quick and try the same procedure with the original system.img
Gerwin_J has quit [Quit: Gerwin_J]
<Turl>
make sure /system is read only then
<Turl>
it should be
<akaizen>
yep it is
<Turl>
then it should work, it may explode though when trying to access the being-rewritten partition
<akaizen>
When I try adb push system.img > /dev/block/system while in recovery it always says target disk ran out of space do you know hwy?
<akaizen>
yea thats fine
<Turl>
you can run 'stop' to kill most of the userland and reduce the chances
<Turl>
I don't think adb push likes block devices much
<Turl>
I was never bothered enough to investigate why either
<steev>
libv: i'd rather not, it's actually quite nice
<steev>
aluminum back
<steev>
in fact, it's one of the nicest allwinner tablets i've seen
<Turl>
steev: probably someone else already identified it for you
<Turl>
steev: google around for brand+model
<steev>
yeah, that's the problem
<steev>
i don't remember
<Turl>
and it's not written on the tablet?
<steev>
and there are like 0 markings, aside from MID970
<steev>
naw
<Turl>
what color is it?
<steev>
silver (aluminum) back, black around the lcd
<akaizen>
fadklsjfhlauefhuwhflkhawf it works... damn wasted a whole week with this shit
<akaizen>
thanks Turl!
<steev>
one sec, i'm booting it up
<steev>
though it definitely doesn't seem to hold a charge as long as it used to
kaspter has quit [Remote host closed the connection]
rz2k has quit [Read error: Connection reset by peer]
deasy has joined #linux-sunxi
ykchavan has quit [Ping timeout: 245 seconds]
HeHoPMaJIeH has joined #linux-sunxi
allen has joined #linux-sunxi
Black_Horseman has quit [Remote host closed the connection]
ykchavan has joined #linux-sunxi
allen_ has joined #linux-sunxi
soul has joined #linux-sunxi
soul is now known as Guest96830
allen has quit [Ping timeout: 260 seconds]
rz2k has joined #linux-sunxi
Guest96830 has quit [Read error: Connection reset by peer]
<steev>
really hard to get excited about allwinner/rockchip considering how hard it is to get decent documentation/source code from them
allen_ has quit [Ping timeout: 256 seconds]
Soru has joined #linux-sunxi
BluesBoy has joined #linux-sunxi
akaizen has joined #linux-sunxi
allen_ has joined #linux-sunxi
<oliv3r>
steev: it's a 2 way thing
<steev>
what is?
<oliv3r>
steev: the A10/A20 have so much potential, if they would have written their drivers properly. We could be busy doing other things
<steev>
that's not true at all
<oliv3r>
So the pure prospect of having a fully GPL endorsable SoC is exciting
<oliv3r>
so it's not that hard to get excited about that ;)
<oliv3r>
the problem is, that the stuff we have is shit
<steev>
fully gpl endorsable?
<oliv3r>
steev: once we are 'done' :p
<oliv3r>
the BROM can be RE'ed if we need
panda84kde has joined #linux-sunxi
<oliv3r>
but other then that, the entire SoC, aside from CAN, GPS, GPU and VPU are now opensource
<oliv3r>
the CAN bus can easily be done or docs obtained for, if we wanted to
<oliv3r>
the GPS can be RE-ed if desired, but incentives for those are very low and they are not really used much or important, so dead silicon I suppose ;)
<oliv3r>
the GPU is lima; the VPU we have something basically done
allen__ has joined #linux-sunxi
<steev>
what gps?
allen_ has quit [Ping timeout: 240 seconds]
<oliv3r>
A10 has a GPS; not sure if they hide it on A20 or actually have replaced it with twi4
n01 has joined #linux-sunxi
massi has joined #linux-sunxi
<andoma>
my dream is some SoC vendor that creates a good documented chip and then ask the community to help development
<focus>
andoma: better still my idea is one of us make success of product, license arm design to make chip, but importantly open sources the entire chip's documentation including video, release one header file that describes the registers and bit fields 100%, and then everyone can jump in and port full linux to it - and never look at any other chips again.
<focus>
The one header file means if the chip design changes, the ONE header file can be changed and hardly any porting required to make all that software work again on new chip.
<andoma>
focus: isn't that more or less what I said?
<focus>
technically yes.
<focus>
but details important
<andoma>
sure
<focus>
if ONE header file describes the chip, the entire embedded world will shift to it
<focus>
sell easily 1 billion chips per year
<andoma>
\o/
rellla has joined #linux-sunxi
vicenteH has quit [Ping timeout: 272 seconds]
<Montjoie>
Hello i have writed the driver for the SS crypto hardware module of SoC A20, and I have random crash (sometimes later after modprobe) does someone want to see it/test it?
<rm>
the driver or the crash
<Montjoie>
the driver
<rz2k>
> SoC vendor that creates a good documented chip - freescale has 5.8K pages ref manual, and many other documents around for i.MX6. community is also involved (freescale guys read meta-freescale Yocto ML, plus there is a community issue system).
<Montjoie>
I work by ssh and I cannot see the jerneldump
<oliv3r>
we have that with the A10, since our hacking makes the documetnation :)
<rz2k>
they also have public git server and they sponsor couple of guys to improve mainline kernel
<rz2k>
andoma: ^
<andoma>
rz2k: that is true indeed
<andoma>
seems their devices are a little more expensive than, for example, aw
allen_ has joined #linux-sunxi
<steev>
not really
allen__ has quit [Ping timeout: 245 seconds]
<rz2k>
I bet that if aw would go freescale way and do all that support\maintaining things, aw devices would end up in the same price range where imx is.
<oliv3r>
Turl: i thought you where doing the SS Crypto driver?
<andoma>
rz2k: yeah perhaps
<oliv3r>
Montjoie: if you don't see kernel dump; then you can't really debug it
<oliv3r>
Montjoie: fi your hacking on the kernel, serial console is more or less a must
<oliv3r>
rz2k: does the freescale have an open bootloader?
<oliv3r>
rz2k: not only the same pricerange, it would be even more popular and used
<oliv3r>
i thought right now, A10 is the cheapest of them all?
<Montjoie>
oliv3r, I have see with turl for the driver
<rz2k>
oliv3r: they are in u-boot mainline, I use 2013.07 u-boot for imx6 devices
<rz2k>
oliv3r: also there's barebox bootloader support, but I dind try it really
<oliv3r>
rz2k: so why did rhombus not choose imx over a10?
Soru has quit [Read error: Connection reset by peer]
<rz2k>
oliv3r: lkcl did imx6 board, but ended not finishing it, dont remember why (guess finishing it was harder than having chinese guys from wits finish the a10/20 board)
<rz2k>
oh, yeah, I forgot about that requirement, somewhere it was said that eoma board will be $30 or something
<rz2k>
so imx is no go here, thats for sure.
<oliv3r>
ah i think imx6 uses vivante
<oliv3r>
how much is the imx soc?
<rz2k>
full featured imx6 with pcie and CAN buses is something like $45, iirc
<oliv3r>
oh wow, costs like 4 to 6 times more then a10
<oliv3r>
yeah taht's quite expensive
<oliv3r>
but does make it an interesting target
<rz2k>
but hey, it has 4 screens output and 4x more cores :p
<oliv3r>
the VPU firmware could eventually be RE'ed aswell
<oliv3r>
but then it's in the same boat as a10, with a10 being just a fraction of it
<oliv3r>
a20*
<oliv3r>
a20 is 7 USD i think
<rz2k>
vivante is an interesting thing
<oliv3r>
yeah it is
<oliv3r>
and it's being RE'd
<rz2k>
you can make it do 600fps in glmark2
<rz2k>
fullscreen 720p
allen_ has quit [Ping timeout: 245 seconds]
<oliv3r>
nice
<rz2k>
sad that they have really weird graphics stack including not only the usual libEGL/whatever stuff, but also ton of their own libs and they use DRI1 (the old one)
<oliv3r>
but at that price, it's simply to expensive
<rz2k>
thats the max featured version
<oliv3r>
the single core basic version is still 22 USD
<oliv3r>
about 3 times the price of a20
<Montjoie>
oliv3r, with serial console, you means using an UART cable ?
<rz2k>
but its nice to have a box that can driver pci-e LTE modem, sata hdd and display a 3d interface at the same time :p
<oliv3r>
Montjoie: yep
<oliv3r>
rz2k: for high performance/servers yeah IMX seems awesome
<rz2k>
s/driver/drive/
<oliv3r>
i mean, with pci-e you can hookup raid controllers etc
<rz2k>
i only tried sierra wireless lte modems and intel wifi cards for now
<rz2k>
but there is gigabit network cards in the list of tested devices
<rz2k>
hope allwinner will place pcie in new A80
<oliv3r>
doubt it
<oliv3r>
i expect a80 to be a31 with diff. arm core
<oliv3r>
maybe updated parts that where bugged
<oliv3r>
little effort, maximal return
<rz2k>
thats supersad, I dont see the point of it then
<rz2k>
where's my killer feature? :p
<oliv3r>
powerVR :p
<andoma>
*shivers*
<oliv3r>
:p
<oliv3r>
a6x might be mali again
<oliv3r>
if not, time for a new soc :p
<rz2k>
agreed
<andoma>
i wonder why they have two different GPU blocks
<rz2k>
time to go for now
rz2k has quit []
<oliv3r>
powerVR performs better
<andoma>
but mali is cheaper to license?
<oliv3r>
i think so
ykchavan has quit [Ping timeout: 272 seconds]
<mnemoc>
moin
<oliv3r>
lo
vicenteH has joined #linux-sunxi
allen_ has joined #linux-sunxi
eebrah has joined #linux-sunxi
eebrah is now known as Guest96730
Guest96730 has quit [Client Quit]
eebrah_ has joined #linux-sunxi
<arokux1>
cellular network is such kind of a mess to me...
<oliv3r>
yep :p
<arokux1>
and all those lock-ins make the situation only worse
<oliv3r>
yeah; though atleast here in NL it's a little easier
<oliv3r>
just to get back to yesterdays discussion about opensource hardware
<lunra>
A wider international citizen-band would have been nice >_>. So that we don't have a situation where there are oligarchs with the exclusive access to radio bands who can charge whatever they like for what is becoming an essential service
<oliv3r>
now that's the proper way to 'use' and 'improve' OSHW
<lunra>
(I understand that exclusitivity of bands is -supposed- to reduce interference, but the result is monopolies)
<arokux1>
lunra: you are talking about some more deeper things
<arokux1>
I am confused merely by the myriad of the acronyms.. :(
<steev>
45 dollars!?
<steev>
maybe if you're buying a single unit
<steev>
or a hand ful of units
<steev>
at 10k + it should only be ~15
<lunra>
hmm, which acronyms? Do you just mean the complicated words? Sorry to those who aren't native English speakers :(
<mnemoc>
most people here is not native english speaking :p
<steev>
wat is dis englis ting
<lunra>
I know, I should have been more careful. Sorry.
<arokux1>
lunra: no! you haven't said anything I haven't understood..
<steev>
mnemoc: most people in here are still better at it than most americans
<steev>
just sayin
<mnemoc>
but to be honest most US citizens probably speak worse english than many non-natives here :p
<mnemoc>
steev: :D
<steev>
gah
<oliv3r>
steev: yeah it's 45 @ 1000 units
<steev>
systemd sucks.
<lunra>
But when I use smaller words among English speakers, they assume I'm being condescending or that I'm stupid ;)
<steev>
sure at 1k units
<steev>
sell harder man!
<arokux1>
what I meant is HSDPA, UMTS, HSPA etc.
<oliv3r>
arokux1: ahh, thought those where pretty 'standard' by now, but isn't athat all '3.5G'
<oliv3r>
with LTE being 4G
<oliv3r>
'4th generation'
<steev>
ugh
<oliv3r>
HSDPA was 3.5G, UMTS was 3G, HSPA i don't remember
<lunra>
Oh, I see. Sorry again ;)
<steev>
at least you specified it means 4th gen, as opposed to the actual 4G speed
<oliv3r>
lunra: stop being so condesending!
<lunra>
wait... what?
<juanfont>
HSPA 3,75 :P
<juanfont>
*3.75G
<oliv3r>
is HSPA faster then HSDPA yeah?
<oliv3r>
or newer*
<arokux1>
you see...
<lunra>
oliv3r: Are you being serious? If so, please tell me where I went wrong, because I did not mean to be condescending.
<oliv3r>
arokux1: never doubted you ;)
<oliv3r>
lunra: i was kidding ;)
<arokux1>
oliv3r: what do you mean?
<wens>
I remember one US carrier marketing HSPA as 4G...
<oliv3r>
arokux1: i agree, it's a maze if you are not 'up to speed'
<oliv3r>
wens: where you the guy working on dvb-* stuff?
<arokux1>
oliv3r: up to speed? I've just decided to dive in :)
<oliv3r>
arokux1: don't have a mobile phone yet?
<arokux1>
oliv3r: how is that connected? I'm happily using my mobile phone just for calling, there is no need to know all those acronyms
<mnemoc>
hahahaa.... just got spam from wits announcing their A20 devkit :p
<mnemoc>
kind of late
<arokux1>
new release or what?
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<mnemoc>
the official $1k A20 development board/kit
cubie has joined #linux-sunxi
<steev>
whoa
<steev>
that's more than the imx6 dev board/kit
<steev>
my sabresd only cost ~650?
<lunra>
Ouch. ~17+ times more than the cubieboard2
<focus>
don't buy it. We bought the $1 A10 kit, were promised linux support, and when it came to the day, they said only Android
<steev>
focus: well what do you expect for $1
<lunra>
hehe
<focus>
$1k
<focus>
hahhahahah!
<focus>
:)
<lunra>
$1 A10 would be nice ;)
<arokux1>
:) so what have they said about their promise?
<lunra>
Perhaps old stock could be obtained at that price one day. Sure will be a lot of unbought tablets :(
<steev>
focus: if it's any consolation, the mx6 has been out for what, 2 years, and they are just now getting hardfloat support for their vivante driver?
<lunra>
What does happen to old unpurchased electronics?
<steev>
lunra: it's not unpurchased
<lunra>
Surely there is stuff that just won't sell? Or does the cost go down until it is finally purchased?
<steev>
it gets shipped to shenzhen and stripped of markins
FR^2 has joined #linux-sunxi
<steev>
no, but uh, depends on the item really
<lunra>
But I suppose some things can't be repurposed
<steev>
if it's something for a military contract, there has to be 10 years worth of supply
<lunra>
LVDS LCDs could likely go into something else, but an actual A10 chip isn't very good in 10 years (except yeah, a military contract)
popolon has joined #linux-sunxi
<wens>
oliv3r: no # 16:46 < oliv3r> wens: where you the guy working on dvb-* stuff?
allen__ has joined #linux-sunxi
<rellla>
did i backread correctly? libv is hacking around mali/mesa? ^^
<lunra>
(if you see this - I really really appreciate your work! Keep it up! :D)
cubie has quit [Quit: Leaving]
allen_ has quit [Ping timeout: 245 seconds]
deasy has joined #linux-sunxi
Soru has joined #linux-sunxi
notmart has joined #linux-sunxi
wingrime has joined #linux-sunxi
allen_ has joined #linux-sunxi
<oliv3r>
rellla: libv is the guy behind lima
<oliv3r>
lunra: yes it is
<Turl>
oliv3r: yeah I was, but I wasn't making active progress so I let Montjoie continue :)
<rellla>
oliv3r: i know :)
allen__ has quit [Ping timeout: 245 seconds]
allen_ has quit [Ping timeout: 272 seconds]
<wingrime>
rellla: whats up with vdpau and xbmc ?
<rellla>
wingrime: nothing so far...not sure about to deal with it, as xbmc seems not to support vdpau+opengles. i'm trying to make the blobs work atm.
<wingrime>
rellla: you can always try use directly our code
<wingrime>
rellla: like in mpeg-sample
allen_ has joined #linux-sunxi
<mnemoc>
libv: did the ssh thing work?
<juanfont>
rellla, any idea on why vdpau+opengles is not supported in xbmc?
fredy has quit [Excess Flood]
fredy has joined #linux-sunxi
<rellla>
juanfont: jemk mentioned - and i think he is right - that the newer versions use the new opengl-feature of vdpau. opengl isn't what we have
<rellla>
wingrime: yes, it would be possible to hack our RE'ed code directly into DVDVideoCodecA10.c
<rellla>
but then we have another program-specific piece of code. doing a more general thing like vdpau would be better
allen__ has joined #linux-sunxi
<wingrime>
rellla: but we have vp8 code thats can't be in vdpau
allen_ has quit [Ping timeout: 272 seconds]
Soru has quit [Ping timeout: 245 seconds]
<Turl>
rellla: libva? :p
<Turl>
you could go the OMX route too
<oliv3r>
OMX is really shit
<oliv3r>
it's a big huge mess, nobody knows really how to properly implement it etc
<oliv3r>
docs are horrible
<rellla>
there would be many ways to reach the goal. we discussed already what would be the most reasonable. so let's start dicussion again, i can't remember how it ended..
<Turl>
oliv3r: but It Just Works! (tm)
<oliv3r>
i think it all needs to be rewritten anyway, so anything that's done now, is just to get things tested
<wingrime>
rellla: we maybe need add more one abstract level for or RE code
<oliv3r>
it needs a completly entirly rewritten kernel driver, and with that, possible vdpau/libva implementation needs to be heavily modified
<wingrime>
rellla: for make same code work without/with vdpau
<oliv3r>
so too much ductape is a little wastefull
<wingrime>
oliv3r: not sure, we still can expose regs using DRM
Sonicadvance1 has quit [Ping timeout: 264 seconds]
<wingrime>
oliv3r: rellla: small libcederus.so
Soru has joined #linux-sunxi
<rellla>
wingrime: building our own libvecore.so upon REing efforts and then use the API for it in various programs (xbmc...) could probably the fastest way to have some success..
<rellla>
wingrime: yes. thought the same.
<wingrime>
rellla: but I not think this is good idea mimic libvecore
<oliv3r>
danger with a small libcederus is, it will become versioning hell; e.g. if we later need one with the proper kernle, things might break horribly :)
<wingrime>
oliv3r: just add version check!
<oliv3r>
:p
<wingrime>
oliv3r: but we have many stuff thats not supported by vdpau
<rellla>
it's the same with mali atm
<oliv3r>
so libvecore.so for now seems not to unreasonable, keeping libcederus.so for the 'proper' version later
<rellla>
kernel<->userspave lib
Sonicadvance1 has joined #linux-sunxi
<oliv3r>
wingrime: also not supported by libva or, god forbid, omx?
<wingrime>
rellla: thats DRM libs
<rellla>
wingrime: i mean keep up right versioning
<oliv3r>
ah well yeah
<oliv3r>
DRM will need to be talked to seperatly obviously
<oliv3r>
but what if we want to do cedarX decoding under the framebuffer?
<wingrime>
rellla: libdrm_sunxi.so required
<wingrime>
oliv3r: ^
<wingrime>
cedar RE stuff maybe need be together with graphics
<arokux1>
I mean is it Tom's or just yet another shop?
allen__ has joined #linux-sunxi
wingrime has quit [Ping timeout: 246 seconds]
\\Mr_C\\ has joined #linux-sunxi
<libv>
mnemoc: great, works :)
<bfree>
libv: building a repository (for one maintainer for debian or ubuntu anyway) is easy, building sane, useful and legal packages to put in there is more challenging ;)
<libv>
bfree: well, i need a slightly hacked mesa, and i will soon provide packages for the lima mesa driver
<libv>
bfree: i expect that ssvb will then also want to prepackage his xf86-video-sunxi
<bfree>
afraid I've had $work help keep me procrastinating away from build a two kernel 3.4 package for A10 and A20 since I got back to my cubies :-/
<mnemoc>
I can build .deb and .yum out of the nightlies if someone documents how :p
<mnemoc>
.rpm
<ssvb>
libv: I'm primarily using gentoo myself, but also somewhat familiar with debian/ubuntu
<libv>
well, np, i can also happily handle the ubuntu side of things :)
<libv>
the biggest mess is going to be keeping mesa recent anyway, as they have no sane api policy in place (on purpose ... which is exactly what i do not like about mesa)
allen__ has quit [Remote host closed the connection]
jemk has joined #linux-sunxi
<bfree>
mnemoc: you should be able to co-opt "my" kernel packaging fairly easily for 3.4 and probably debian's for 3.10 (and maybe mainline) ... but probably better to wait until I get it building multiple flavours anyway :-/ and of course, then there is the question of the config, e.g. my 3.4 is very modular while I get the impression more stuff built in would be the preference for you guys :-/ and e.g. debian's 3.10 is armmp and you might not want to be ...
<bfree>
... annoyed by non-sunxi issues :-p u-boot is probably easier
<ssvb>
bfree: building legally sane packages is a bit tricky, r3p0 blob is a single shared library libMali.so, which makes it easier
<ssvb>
bfree: but r3p2-01rel2 is a set of blobs, which are also a bit buggy
<libv>
don't bother packaging the blob anymore :p
<ssvb>
libv: what about the shader compiler?
<mnemoc>
bfree: I would like to integrate it into the currently nightly building script. no patches
<libv>
we should fix that over the next few months
<libv>
people have been happy with the direct installation
<libv>
and things should not be made too easy and comfortable for the binaries either
<bfree>
mnemoc: what do you mean by "no patches"?
<ssvb>
libv: ok, no blob dependencies is very nice
<mnemoc>
bfree: not generating a patch against a tarball but directly out of the git working dir
<mnemoc>
bfree: and using the same files already compiled
Gerwin_J has joined #linux-sunxi
<rellla>
juanfont: i think so. but i always mix this gl/gles/mesa ... things up. so it's not clear for me atm, how things work together or can work together in xbmc. libv will know what will happen with lima and mesa and what we maybe can do with it in future...
<libv>
as a user, you should notice no difference
<libv>
apart from it being 6% faster than the binary (in the very general case) and it is built externally
<libv>
rumour has it that you never get even close with gallium on performance
<oliv3r>
rellla: GL is OpenGL (the full thing) GLES is Embedded Systems, a reduced GL
<libv>
but that's not my motive for going with the intel glsl compiler: gallium would not have allowed me to bring up the command stream separately
<oliv3r>
libv: i heard you are looking for devs for a graphics-dev room? :)
<libv>
oliv3r: you're already having a talk, don't overload yourself :)
<libv>
but yes, google should be able to turn things up on "graphics devroom"
<oliv3r>
libv: oh i have no intention :p
<oliv3r>
libv: and I am? :p
<oliv3r>
IF i am going, i deff. wanna be IN the room listening ;)
<libv>
there should be campus wide video recordings this year
<libv>
even in the devrooms
naobsd has joined #linux-sunxi
<libv>
one of the fosdem organizers is hard at work organizing that now
eebrah has joined #linux-sunxi
eebrah is now known as Guest38016
<oliv3r>
awesome
<oliv3r>
when willt he normal talks be announced?
Soru has joined #linux-sunxi
<libv>
oliv3r: soon iirc
Guest38016 is now known as eebrah_
vicenteH has quit [Ping timeout: 246 seconds]
JohnDoe_71Rus has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org]
ykchavan has joined #linux-sunxi
<ykchavan>
wiki does not support 'bgcolor' tag. So, can not mark individual field with colors.
<libv>
and the amount of money burnt on that meant that i, as one of the few people of that company who was making constant money for that company, am now no longer making any money.
<libv>
it was existing hw, that never was solid enough to be sold
<ganbold_>
I see
<panda84kde>
libv: is Codethink in trouble?
<libv>
it was
<libv>
now they are on a hiring spree again
wingrime has joined #linux-sunxi
FDCX has joined #linux-sunxi
<wingrime>
jemk: your last additons about interlanced?
<jemk>
wingrime: what last? the scaling lists are nothing about interlaced, some bits i added earlier are for interlaced, but it needs more work
BluesBoy has quit [Quit: HydraIRC -> http://www.hydrairc.com <- The professional IRC Client :D]
<arokux1>
hipboi: I do not know is this is a secret info, but how many cubieboard have you sold already? just curious...
<hipboi>
arokux1, it's not secret
<hipboi>
arokux1, more than 20k
<hipboi>
that's not very many
<arokux1>
if you compare it with RPi, yes
<arokux1>
they've sold 1.75M units
<hipboi>
but i am happy with it
<arokux1>
hipboi: are you worried that lots of companies decided to make their boards?
<arokux1>
hipboi: i thought it is somehow affiliated with you
<arokux1>
"All stuffs are quickly shipped from our factory in China directly to you."
<arokux1>
so this is a lie...? ----^
<ssvb>
hipboi: is there a significant percentage of cubieboard sales in China compared to the rest of the world?
<hipboi>
it's not on the cubieboard.org/buy...
<hipboi>
ssvb, at first 80% cubieboards are sold outside China
<popolon>
so about 4K in china
<ssvb>
it's interesting, I guess that anyone in China would normally prefer Cubieboard over Raspberry Pi, that's a rather nobrainer choice :)
<ssvb>
now I wonder if Raspberry Pi is actually known/popular in China
<arokux1>
RPi has huge community and lots of shields..
<arokux1>
cubieboard is just a SBC
vicenteH has joined #linux-sunxi
Soru has quit [Read error: No route to host]
Soru has joined #linux-sunxi
ganbold__ has joined #linux-sunxi
<wingrime>
ssvb: can you review last patrick patchset
ganbold_ has quit [Ping timeout: 268 seconds]
<wingrime>
ssvb: I not sure about DMA coherent, and cache stuff
<ssvb>
wingrime: "dmacoherent" is mostly just the alias for writecombine or uncached
<ssvb>
wingrime: writecombine means fast buffered writes and slow uncached reads
<ssvb>
wingrime: uncached means both slow reads and writes
<wingrime>
ssvb: please send mail to ML, about, it looks fixable in userspace
<wingrime>
ssvb: but I not want make any regression
nove has joined #linux-sunxi
<ssvb>
wingrime: depends on the userspace (if the userspace is the allwinner blob, then it's not really fixable)
servili007 has joined #linux-sunxi
<wingrime>
ssvb: there wrapper code that calls ioctrls
<wingrime>
ssvb: blob not call ioctrl directly
<ssvb>
wingrime: the blob may call a bunch of callback functions, but I have not really looked at them in detail
<nove>
wingrime, jemk, i just got the first partial h264 encoder trace
<nove>
43 registers from a00 to bc4
pfdm has joined #linux-sunxi
<pfdm>
morning
<jemk>
nove: nice, so you managed to run the demos?
<pfdm>
I'm trying to understand the new blob mem allocator options
<wingrime>
nove: cool
<nove>
jemk, the enc_dec_test binary as is, without compiling it
<pfdm>
my kernel lack sunxi-mem , someone try to make them work with the cdxalloc lib
<pfdm>
?
<jemk>
nove: ah, ok. i have some self compiled version here that runs, but it produces garbage :(
Soru has quit [Read error: Connection reset by peer]
Soru_ has joined #linux-sunxi
jemk has quit [Ping timeout: 272 seconds]
n01 has quit [Ping timeout: 268 seconds]
<nove>
pfdm, cedar related?
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
servili007 has quit [Read error: Connection reset by peer]
panda84kde has quit [Quit: Konversation terminated!]
<pfdm>
nove: yes
<nove>
pfdm, i can not help, my understanding of kernel side is not very good
<pfdm>
well it's most on the libcedarv side. Looks like they change the memory allocator, but i can't see where they call (or I should call) cedarx_hardware_init
<pfdm>
maybe I can ask rellla
hipboi has quit [Quit: Leaving]
<wingrime>
techn_: thats not fix
<wingrime>
techn_: disabling cached access make it SLOW
<wingrime>
techn_: issue present only on cortex a7 due cache system difference
<wingrime>
techn_: thats in blob
massi has quit [Quit: Sto andando via]
<wingrime>
techn_: it's better fix in userspace without regressions , or limit a20, a31
<ssvb>
techn_: "Consistent memory is memory for which a write by either the device or the processor can immediately be read by the processor or device without having to worry about caching effects."
<ssvb>
techn_: and there is also "dma noncoherent" mapping, which means that cache flushing has to be done explicitly
<ssvb>
techn_: apparently the cedar driver has been originally designed to use cached memory and cache flush ioctl, but the userland code does not properly play by the rules
<ssvb>
techn_: and I guess they decided that it was easier to just change the memory mapping for a20 on android instead of debugging it properly :)
akaizen has joined #linux-sunxi
<akaizen>
Is there an A20 u-boot for NAND and android?
<techn_>
ssvb: but wouldn't that dma coherent memory solve that issue? and after thinking.. isn't that issue.. memory must be dma coherent
<ssvb>
techn_: if we make it dma coherent, then the memory corruption issue is solved and we don't need the cache flush ioctl anymore
<techn_>
and memory is still cache'd?
<ssvb>
no
<ssvb>
the "dma" buzzword and the terminology from the kernel framework layer does not magically make the hardware behave better
jemk has joined #linux-sunxi
<wingrime>
oliv3r: a20 is sun7i?
<akaizen>
wingrime: yes
deasy has joined #linux-sunxi
ZetaNeta has joined #linux-sunxi
ykchavan has quit [Quit: Leaving]
pfdm has quit [Quit: Page closed]
mcbrick has joined #linux-sunxi
<ssvb>
techn_: still dmacoherent mapping is the same as writecombine in the sunxi kernel (plus L_PTE_XN bit), so it's a reasonable choice is we never read back the decoded data using the CPU
<techn_>
ssvb: I just was thinking that it would make software flush later.. automaticly.. but ok
ykchavan has joined #linux-sunxi
<wingrime>
ssvb: we have encoder, and some return bit's from cedar
<wingrime>
ssvb: thats imposible not read anything from cpu
<ssvb>
wingrime: the performance of uncached reads can be also improved if they are done in a special way
<ssvb>
wingrime: but just a naive C loop or ordinary memcpy will be slow
<wingrime>
ssvb: better simply flush cache once for each video frame, not wait ddr3 for each access
ykchavan has quit [Ping timeout: 268 seconds]
<wingrime>
ssvb: can you post something in thread with cedar patchset (for partic)
ccube has joined #linux-sunxi
<libv>
bfree, mnemoc: do we have a sunxi packaging gpg key?
<jemk>
I still believe that the cache issues can be solved in userspace. The blob calls IOS.mem_flush_cache() where necessary(), which has to be implemented by higher layer
<libv>
this would be used to create signed packages and a signed repository
<libv>
shortly before stuffing packages in the repository, one would run a re-sign on the relevant packages
<ssvb>
jemk: in android too?
<jemk>
ssvb: in android it works as far as i know
<jemk>
ssvb: there the higher layers are implemented correctly
<jemk>
if i would have my ct already i would test it, but i'm pretty sure
<ssvb>
jemk: hmm, maybe I misunderstood Patrick, but it seemed to me that he mentioned that Android is not having any issues because the android kernel is using dmacoherent mapping
<oliv3r>
wingrime: of course :p a31 is sun6i
<ssvb>
jemk: I believe you can just enable write allocate cache on A10 to run the tests
<wingrime>
jemk: post to ML
<wingrime>
ssvb: cedarx report same Version for a10 and a20, so old blob with OLD cache behavior should works for a20 but with cache issues
<ssvb>
jemk: write-allocate cache is just slower on Cortex-A8, but other than this it is perfectly supported too
<jemk>
ssvb: ok, i will try
focus_it has joined #linux-sunxi
<wingrime>
jemk: not cedarx have same id in version register for a10 and a20, so old blob with OLD cache behavior should works for a20 but with cache issues
<wingrime>
*notice
<lkcl>
rz2k: yes basically we need someone to pay for finishing the iMX6 board, or someone to do a deal with in getting it into production.
arokux2 has joined #linux-sunxi
<arokux2>
lkcl: while you are here.... when will we get EOMA?! :p
<wingrime>
lkcl: better EOMA with zync-7000
<wingrime>
err
<oliv3r>
lkcl: hey! how are things?
<arokux2>
wingrime: why it should be EOMA?
<wingrime>
Zynq-7000
<wingrime>
arokux2: two core cortex a9 with internal FPGA
<wingrime>
arokux2: and PCIe
<arokux2>
wingrime: yeah, but why it should be EOMA, go for ZedBoard
<oliv3r>
i need to get a DP open logic bench sniffer
<wingrime>
arokux2: digital radio reciver/transmitter
<jemk>
nove: i compiled with some old armel compiler and linked the static lib, and after adding some unknown symbols by hand it worked but the resulting h264 was corrupted, so i gave up again
<wingrime>
jemk: looks cache problem?
<arokux2>
wingrime: redpitaya is Zynq based
<wingrime>
arokux2: yes, i know
<wingrime>
arokux2: there many things can done with fpga
<jemk>
ssvb: thanks, "works" ;) now i can check my theory
<arokux2>
wingrime: so you think you can beat redpitaya price tag?
<lkcl>
arokux1: when the MEB's been done!
<arokux2>
lkcl: you should prepare things in parallel!
<lkcl>
wingrime: i'd looove to see a board with one of those. xilinx however.. well... they price things based on expected volume...
<lkcl>
arokux2: i did - nobody's taking anything up! i've done an ingenic jz4760 cpu card, an imx6 cpu card...
<lkcl>
i think once we have a first cpu card and first product out, it will go nuts.
<wingrime>
lkcl: fpga very intersed thing
<arokux2>
lkcl: what about the A20-EOMA? there are lots of people waiting...
<wingrime>
arokux2: just wait when pre-order ends and we see real price
<lkcl>
because we will be able to fund these things without relying on other people, and also it will no longer appear to be vapourware
<lkcl>
wingrime: you and me both! we're not the critical path any more, the MEB is the critical path
<lkcl>
hey guys i'm really sorry i have to get back to STM32F103 firmware. i *have* to get this camera working. then we can sell 10k tablets (yay!)
<arokux2>
lkcl: spread the cpu card - people will do lots of the work for you
<lkcl>
arokux2: that's what's happening - indirectly. it's a narrow bottleneck at the moment, but the first 2.5k order once that's out (with a MEB on every card) it'll go nuts after that.
<wingrime>
lkcl: first step always hard, but now you can change world
<wingrime>
lkcl: i hope))
<lkcl>
wingrime: i can *already* change the world.... i'd like it to be a bit better thanks :)
<arokux2>
lkcl: what is the "narrow" bottleneck right now?
<arokux2>
lkcl: who are we waiting for? :p
<lkcl>
arokux2: only one CPU Card and we're waiting for the MEB.
<lkcl>
arokux2: i did say that (3 times) now! the MEB is the bottleneck.
<oliv3r>
lkcl: any contact with AW latley? or that all still happening in the background
<lkcl>
oliv3r: not recently.
<oliv3r>
:(
<arokux2>
lkcl: will you still do this router thingy?
<wingrime>
lkcl: I want see h264 example in hard-float build, it's requred for make traces and make RE
<wingrime>
lkcl: are you tested libvdpau-sunxi?
<lkcl>
arokux2: yes! i found the right chip (and datasheet)
<lkcl>
wingrime: no. too busy.
<lkcl>
hey guys i reaalllly have to get on, i have very limited time.
<oliv3r>
lkcl: go fix that firmware1
<oliv3r>
lkcl: STM32 is a microcontroller, ist hat used for the firmware?
<lkcl>
evenings only, for about 1.5 hours, and i *have* to get this STM32F103 firmware to the point where i know for sure the camera data can be read.
<oliv3r>
that's quite a overrated micro, arm-m0
<lkcl>
oliv3r: yes. it's for the speakers (done, working), backlight (doable), microphone (done... sort-of), and camera (*really* pushing the limits)
<wingrime>
lkcl: eh, there is this tablet proto photo?
<lkcl>
wingrime: that's the one. with a huge bristling set of resistors sticking out of it :)
<oliv3r>
lkcl: so you use it as an Embedded Controller; very cool
<wingrime>
lkcl: link
<oliv3r>
lkcl: ah you can't use the PWM for backlight
<wingrime>
lkcl: and why not use stm32f4 ?
<lkcl>
oliv3r: yes. the alternative was $1.50 for a USB-audio IC, $2 for a USB-camera IC, this, that and it can all be done with a $2.36 72mhz Cortex M0
<lkcl>
wingrime: because the STM32F4 is $5 not $2.36
<wingrime>
lkcl: f4 probably can play music without cpu on lowpower
<oliv3r>
lkcl: who's writing the firmware?
<lkcl>
wingrime: indeed. but so can the stm32f103.
<lkcl>
oliv3r: me.
<oliv3r>
ah, ok; go go go then ;)
<lkcl>
i'm doing the preliminaries and you're stopping me from doing it!! :)
<oliv3r>
lkcl: hushy! go!
<lkcl>
lkcl@teenymac:/opt/stm32f/eoma-firmware/eoma68/st_port103r/STM32F103-DualCDC/Project/DualCDC/src$ ls
<nove>
wingrime, i mean, encoder libraries with a linux_lib directory
<ccube>
hey! :) I got some strange happenings here and need to figure out if this is an HW or SW issue. The base is a A10s-OLinuXino-M. I have got several of them here and all are working fine with my custom u-boot/kernel compilations. So, now there is a complete customizes version of this board produced for us. I got some of the first prototypes to test my software on it, but I cannot go with it. With some older u-boot versions I am always getting a CRC error when
<ccube>
Verifying my kernel checksum. No I made a new u-boot from current git. This sais: "Verifying Checksum ... OK", then loads the image, but goes away with an kernel oops immediately. every boot another error address. :( May this be a configuration issue, or seems my hardware to be faulty?
<ccube>
(cross posting in #olimex)
<nove>
wingrime, is alook alike the olimex enc_dec_demo, but with *include* *linux_lib* directories
<oliv3r>
ccube: post log, can't smell the oops :)
<oliv3r>
ccube: otherwise we don't know what you did wrong
<ccube>
same sd card in a10s olimex board is booting fine. the "faulty" board is missing some of the hw-components ( no audio, no hdmi, no second sd slot, no gpio, no usb host). dont know if this matters
<ccube>
it also happens that everything freezes completely
<ccube>
so everything completely unreproducable, but it realle never boots to userspace :(
<oliv3r>
did you disable all missing hardware in your fex file?
<jemk>
wingrime: i have made a fixed libcedarx that works without disabled cache.
<oliv3r>
ccube: sysfs_find_dirent is failing with a strcmp
<ccube>
i think i did
<ccube>
for me it looks like a memeory issue soimewhere
<oliv3r>
no
<oliv3r>
it can't find the sysfs entry
<wingrime>
jemk: flush cache realy called?
eagles0513875 has quit [Ping timeout: 272 seconds]