<spv>
If you're referring to the big.LITTLE architecture part, that's just for power savings
<spv>
only 4 cores are active at a time.
<wingrime>
cool
<Wizzup>
well, no
<Wizzup>
On some you can use all 8
<Wizzup>
afaik
<spv>
I was under the impression the A7's replace the A15s for when you want to go into a lower-power mode
<Wizzup>
yes, but you can also enable them at the same time
<spv>
or A8s
<Wizzup>
on some at least :)
<spv>
I believe I saw information on the ARM website about the two being mutually exclusive, but my memory is bad.
<spv>
in more than one way: I'm unable to write to memory on my A20
fredy has joined #linux-sunxi
<wens>
wikipedia mentions it depends on scheduler support
<spv>
kk
<Wizzup>
No, I think some exynos could, others could not
<Wizzup>
buZz: ping
<spv>
jobs
<spv>
mt
<wens>
I think at first it was either 4 A7 or 4 A15 active
<wens>
then came 4 A7-A15 pairs that could be individually switched
<wens>
hoping for heterogeneous multi processing though
<mnemoc>
exynos 5410 wasn't able to use the 8, and had a PVR GPU, exynos 5420 fixed both mistakes :p
<Wizzup>
lol
<mnemoc>
heterogeneous multi processing is probably the big difference between their newly invented sun8i and sun9i machs
<wens>
wasn't sun8i a7-smp?
<mnemoc>
they regrouped all their a7-only in the sun8i bucket
<mnemoc>
and sun9i talks about a15 in Kconfig
<gzamboni>
spv yes, i just checked the specs , in the video she says they can run the 8 cores simultaneosly at the same time
<mnemoc>
but if a20 and a31 and a24 are all the same, the difference with sun9i would be the capability of using the heterogeneous cores at the same time
<mnemoc>
the doubt is still mail t6xx vs. PVR6
<spv>
Speaking of smp, anyone working on A20 SMP?
<wens>
wasn't that solved with PSCI (PCSI?), but a few bugs maybe in u-boot?
<spv>
Not certain -- looking.
<spv>
what I know thus far is only 1 CPU comes up with a SMP kernel.
<spv>
and most of the platform smp code looks like it's for the A31
<wens>
mnemoc: i don't think i would use it much, and bandwidth is not a problem for me :)
<mnemoc>
wens: it also helps for ssh tunnels under the wall :p ... but cool. if you ever need an account there don't hesitate to ask
<wens>
i'm not in the wall :p
<mnemoc>
=)
<mnemoc>
oh, I see you are from the beautiful (formosa) island
<wens>
:)
<wingrime>
uboot have strange variants of sunxi board
<wingrime>
sun8iw3 ?
<mnemoc>
wiki wiki wiki
<mnemoc>
wingrime: see linux's Kconfig
<mnemoc>
they redefined the sunNi
<mnemoc>
and unified a lot
<NightShade>
anyone got a working fw_env.config for the user space u-boot tools with u-boot on mmc?
<mnemoc>
wens: maybe you can help us getting a mirror of dl.linux-sunxi.org over there :p
<NightShade>
I have checked the wiki :)
<wens>
mnemoc: not enough space atm :(
wolfy has joined #linux-sunxi
<mnemoc>
:(
<hramrach>
hello
<hramrach>
wingrime: some guy AMD made the 8bit UVD on Radeons decode 10bit video
<mnemoc>
wens: connectivity and bw is harder and far more expensive than storage, and you already have those ;-) ... so only a minor problem remain
<hramrach>
so maybe you want to look - it's implemented as OpenMAX support inside Mesa which nothing I am aware of can use
<wens>
mnemoc: i'm living off university servers. i was the sysadmin for my department when I was at school
<wens>
i'll look into doing a mirror on the departments mirror server
<hramrach>
from the looks of it it converts the 10bit stream into 8bit stream on the fly and then feeds it to the hw decoder
<wens>
since we already have debian, and adding ubuntu atm, adding a bit more opensource stuff shouldn't be a problem
<mnemoc>
wens: we have rsync already running
<hramrach>
speaking of mirrors ..
<hramrach>
I guess I Should switch to Ubuntu since you can make Ubuntu PPAs and have them built automagically
<wingrime>
hramrach: wow
<hramrach>
running Debian is fine but extending it is a real pain
<wingrime>
hramrach: but, I note that 10bit video not by H264 standart
<wens>
you can still use PPAs from debian, no?
<wens>
dependencies might be screwed up a bit
<hramrach>
wens: yes, you can. But for the ppa to be really useful it should be built on the same set of libraries as the target system you are installing to
<mnemoc>
libv also maintains packages.linux-sunxi.org/debian, not only /ubuntu
<hramrach>
wingrime: standard or not it's what people use to encode video quite a bit
<hramrach>
encoding 10bit and decoding 8bit is actually quite a bit worse because you trash 20% of the data you have and you probably truncate, not round
<hramrach>
but with the conversion you can decode the video without resorting to re-encoding it to even preview it
<hramrach>
*quite a bit worse than having an 8bit video to start with
<hramrach>
I built Mesa 10 with patchset on top which is supposed to implement this but I have no player that can use the stuff so I don't even have any idea if it works
AreaScout has joined #linux-sunxi
<hramrach>
gstreamer supposedly can use omx but the only gstreamer media player I know it totema and that is made by gnome which means it's totally unusable bugpile
<hramrach>
*totem
<wingrime>
mnemoc: aw in own style , libisp blob in kernel
<wingrime>
libisp.ar
<mnemoc>
wingrime: send the request to Eva please
<mnemoc>
they gave the libnand.a sources to lkcl when asked
<wingrime>
lkcl: ping
<mnemoc>
eva.wu allwinnertech com
quintin_ has joined #linux-sunxi
<mnemoc>
try to be friendly and assuming it was a naive mistake, not ill intentioned
<wingrime>
mnemoc: I simple no idea, is that same ISP that in cedar or something also
<wingrime>
mnemoc: better ask lkcl
<wingrime>
I never done such thing before
hipboi has joined #linux-sunxi
<mnemoc>
wb hipboi!
<mnemoc>
hipboi: btw, what's the name of the storage technology you use in the radxa rock? I remember it was something similar to emmc but not emmc
<hramrach>
look at the rk flyers
<mnemoc>
i was wondering if the a80 board could be using something similar
<hramrach>
hmm, actually they updated the web to say eMMC now
<hramrach>
basically by the looks of it it was eMMC without paying for the trademark
<mnemoc>
:p
<hramrach>
iNand :p
<hramrach>
but the one on RR had yet another name iirc
<hramrach>
there should be specs
<hramrach>
for the hw parts somewhere
<wingrime>
looks we now how one wire hw controller
<wingrime>
*have
<hramrach>
I had it for ages on my ancient AGP MGA card and never used it ;-)
<hramrach>
but good for people with some cool 1w devices I guess :)
<mnemoc>
wingrime: yes, there is a 1w driver in 3.4 :)
jemk has joined #linux-sunxi
nove has joined #linux-sunxi
<wingrime>
mnemoc: cool news for you
<wingrime>
mnemoc: we now have DMAENGINE driver!!
<wens>
really?!
<hramrach>
mnemoc: 2013-06-03 13:16:28 hipboi called tSD
<jemk>
nove: is the description of ve reg 0xb04 from jpeg? because it doesn't fit to what i found in h264, it looks like we have dual use registers again.
<nove>
jemk, yes from jpeg, it sets 0x8 and 0xc in the high bits, and also h264 sets 0x8,, this could mean some function that defines the meaning of the lower bits
<nove>
jemk, but you right, is not the naming, those bit fields should have that name
<nove>
jemk, those bit fields should *not* have that name
<wingrime>
jemk: are you sure 0bxx are for jpeg too, I thinked 0xa for jpeg???
<nove>
jemk, i am clueless for names, so fell free to name them as fitted
<oliv3r>
not sure if osadl is what we need; but the wiki page isn't that clear
<jemk>
nove: ok, then I will add a second table with its h264 meanings.
<nove>
wingrime, jpeg(mpeg) uses the same subengines as h264
<wingrime>
nove: so, ISP looks only alpha mix, resize?
<jemk>
wingrime: isp is source for avc it seems, as avc doesn't have any input related regs
<oliv3r>
mnemoc: smtp-auth yes, but without ssl I think;
<nove>
wingrime, its looks like, and yes what jemk said
<wingrime>
jemk: so, isp resize camera output to encoder size
<wens>
dmaengine driver looks good at first glance
<andhe>
oliv3r: maybe spi-inc.org can help you if you want to be able to receive money without all the administrative hassle.
<wens>
add devicetree and individual soc support, and should be good
<andhe>
oliv3r: (spi holds debians assets)
<wingrime>
wens: check also, ac97 audio driver
<mnemoc>
oliv3r: we can kidnap the cert from nginx or configure nginx as reverse smtp proxy
<wingrime>
jemk: we can check ISP using vpdau or so
<wingrime>
jemk: for smple resize
<oliv3r>
why eMMC is 'easier' for us to work with now, i still pray for an MTD driver, as that would make the chip even more open (eMMC has a propriatery micro, with nand we don't have that)
<mnemoc>
all your your eMMCs are belong to us! -- NSA
paulk-collins has quit [Remote host closed the connection]
<libv>
operation complete :)
<libv>
now some more wiki work and then u-boot
<mnemoc>
tablet survived?
<libv>
the patient was fully conscience all the time :)
<wens>
andhe: fw_bcm40181a2*.bin and nvram_ap6210.txt
ZetaNeta has joined #linux-sunxi
ZetaNeta has quit [Changing host]
ZetaNeta has joined #linux-sunxi
wingrime has quit [Ping timeout: 272 seconds]
<andhe>
wens: thanks!
<wens>
andhe: I can't find bcm40181 in official broadcom brochures though
pacopad has joined #linux-sunxi
ZetaNeta has quit [Ping timeout: 252 seconds]
quintin_ has joined #linux-sunxi
n01 has joined #linux-sunxi
ZetaNeta has joined #linux-sunxi
LoCoZeNoz_ZUE has joined #linux-sunxi
LoCoZeNoz_ZUE has joined #linux-sunxi
rz2k has joined #linux-sunxi
LoCoZeNoz_ZUE has quit [Quit: Leaving]
<nove>
the cedar encoder binary libs from the A23 sdk, gives the same problems as i had when i tried the ones from A31
tomboy65 has joined #linux-sunxi
tomboy64 has quit [Ping timeout: 265 seconds]
Quarx has quit []
morfoh has quit [Quit: leaving]
jemk has quit [Remote host closed the connection]
<Seppoz>
happy 2014
paulk-aldrin has joined #linux-sunxi
fredy has quit [Excess Flood]
fredy has joined #linux-sunxi
Nyuutwo_ has quit [Ping timeout: 245 seconds]
<lkcl>
wingrime: yep. a not-exactly proper version, clearly the top of a git repo at the time, but yes.
<lkcl>
mnemoc, wingrime: and there are other instances where it's ended up being released as well.
<lkcl>
i'm sure it's in cb2-dev 3.4 for example
<pacopad>
Hi all i got a compilation problem on my cbé, for exemple i try compile the test provided by sunxi-mali
<pacopad>
i need to add -lX11 in the makefile
<pacopad>
i compiled this soft before without adding that
<mnemoc>
lkcl: this is about the A23 and A31/A31s SDKs... cb2 uses A20
<mnemoc>
with the A20 we have everything more or less sorted out
morfoh has joined #linux-sunxi
<lkcl>
mnemoc: ah ok.
tomboy64 has joined #linux-sunxi
<mnemoc>
and they will launch their own board with A80 (4xA7+4xA15) in CES
<mnemoc>
at*
tomboy65 has quit [Ping timeout: 252 seconds]
fredy has quit [Excess Flood]
fredy has joined #linux-sunxi
AreaScout has quit [Ping timeout: 272 seconds]
TehCapta1n has joined #linux-sunxi
CaCtus491 has joined #linux-sunxi
CountryGeek has joined #linux-sunxi
bfree has joined #linux-sunxi
torbenh4 has joined #linux-sunxi
aep_ has joined #linux-sunxi
speakman has joined #linux-sunxi
speakman has joined #linux-sunxi
speakman has quit [Changing host]
tomboy64 has quit [Ping timeout: 240 seconds]
menomc has joined #linux-sunxi
hurtigbuffer has joined #linux-sunxi
jelly-home has quit [Disconnected by services]
hurtigbuffer is now known as jelly-home
Fangorn has quit [*.net *.split]
speakman_ has quit [*.net *.split]
aep has quit [*.net *.split]
torbenh3 has quit [*.net *.split]
bfree_ has quit [*.net *.split]
CaCtus492 has quit [*.net *.split]
TehCaptain has quit [*.net *.split]
dwilkins has quit [*.net *.split]
mnemoc has quit [*.net *.split]
menomc is now known as mnemoc
CountryGeek is now known as dwilkins
Nyuutwo has joined #linux-sunxi
[7] has quit [Disconnected by services]
TheSeven has joined #linux-sunxi
book`_ has joined #linux-sunxi
oliv3r_ has joined #linux-sunxi
tkoskine_ has joined #linux-sunxi
wolfy has quit [Ping timeout: 252 seconds]
codyps1 has joined #linux-sunxi
oliv3r has quit [*.net *.split]
book` has quit [*.net *.split]
arete74 has quit [*.net *.split]
codyps has quit [*.net *.split]
tkoskine has quit [*.net *.split]
arete74 has joined #linux-sunxi
iamfrankenstein has quit [Quit: Nettalk6 - www.ntalk.de]
fredy has quit [Excess Flood]
fredy has joined #linux-sunxi
<libv>
aha, u-boot.
<oliv3r_>
sweet
<oliv3r_>
did oyu have to change alot?
<libv>
oliv3r_: first off, i am working on a freshly acquired a20 tablet atm
<libv>
i am not playing with the a23
<libv>
and no, the changes are pretty standard
wolfy has joined #linux-sunxi
book`_ has quit [Ping timeout: 252 seconds]
n01 has quit [Ping timeout: 252 seconds]
Black_Horseman has joined #linux-sunxi
<oliv3r_>
yeah i knew you put the a23 to the side for the timing beeing
book` has joined #linux-sunxi
<libv>
has anything ever been done for the .fex to devicetree translator?
Gerwin_J has joined #linux-sunxi
<oliv3r_>
yes
<oliv3r_>
mripard: and mnemoc
<libv>
ah nice, i hope that that doesn't get forgotten, as we really are in danger of losing our biggest advantage
<libv>
oliv3r_: be sure to point that out in your fosdem talk: we can bring up a proper linux on a previously unsupported device quickly and with little hassle
<libv>
we are not limited to development boards :)
<mnemoc>
well.... my fex to dtb id static, merely an extra output format for `fexc`. mripard is writting a dynamic layer between the legacy bootloader and mainline kernel to translate on the fly....
<mnemoc>
so I output dts, mripard wants to output dtb
<libv>
good to know :)
vector80 has joined #linux-sunxi
book` has quit [Ping timeout: 240 seconds]
book` has joined #linux-sunxi
<netlynx>
oliv3r_, in what devroom/track is your Fosdem talk gonna be ?
<libv>
netlynx: mainline, in the K building
<libv>
iirc
rah has quit [Ping timeout: 245 seconds]
<oliv3r_>
so one solution is short term, the other longer term
<oliv3r_>
kk
<oliv3r_>
but yeah, the fex file should def. gain some mentioning
<gzamboni>
apparently aw is testing some already mainlined drivers. in the last 3.4 A23 source in /pinctrl/src/sunxi_pinctrl_test.c they have the mripard code with an AW email contact: * Shaorui Huang <huangshr@allwinnertech.com>
<gzamboni>
they have also the pinctrl mainlined sunxi driver backported in this 3.4 kernel
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<mrnuke>
oliv3r: I see. BTW, what parts need refactorting? If I'm going to rewrite something for corebooy, maybe you guys can make use of it as well
<oliv3r>
mrnuke: dram.c is the biggest candidate right now
<oliv3r>
biggest problem there is, it's so fidety, we dno't know what has to initted when
<mrnuke>
oliv3r: ever did DDR3 init from scratch before?
<libv>
so i fear we might get stuck with a pvr rogue
<oliv3r>
mrnuke: nope :)
<oliv3r>
libv: i think so; check amery's google+ feed
<oliv3r>
libv: benn kinda hinted it's a cubietech design
<oliv3r>
libv: it may be mali t6*
<mrnuke>
oliv3r: it ain't pretty. DDR3 has mode registers, that you need to access and configure before even thinking of calibrating a capture window and three or four sets of delays
<mrnuke>
oliv3r: luckily, A10 seems to do most of that automatically (also not uncommon for DDR3 controllers)