<wens>
I was setting the pin for clk_out in pinctrl-1
<wens>
turns out it wasn't active
<Turl>
wens: ah
<Turl>
you need another , "defalt" on -names to do that I think
<Turl>
each name there is a -N iirc
<wens>
putting all of them in pinctrl-0 works as well
<wens>
and won't 2 "default" names collide?
<Turl>
I never tried tbh
<Turl>
but I believe the nth name on the list triggers a scan of pinctrl-N as to say
<libv>
hrm, wrong uart set and no early printk.
<wens>
well if you don't give it a name, it becomes a second pinctrl set (state)
<libv>
ok, now .75 does boot, but this default uart thing is pretty nasty
<libv>
it must also touch some other bits when it defaults to uart1, as it never enabled disp otherwise
[7] has quit [Read error: Operation timed out]
TheSeven has joined #linux-sunxi
pwhalen has quit [Read error: Operation timed out]
* libv
gives himself a thorough kicking
<libv>
sd-cards work better when they actually have a script.bin on them
pwhalen has joined #linux-sunxi
pwhalen has joined #linux-sunxi
pwhalen has quit [Changing host]
<libv>
well, at least i cleaned up some bits on the wiki, again.
<libv>
aha, it doesn't know the nand
Gerwin_J has quit [Quit: Gerwin_J]
Black_Horseman has quit [Remote host closed the connection]
VargaD has quit [Ping timeout: 284 seconds]
bfree has quit [Ping timeout: 284 seconds]
bfree has joined #linux-sunxi
dlan_ has quit [Ping timeout: 272 seconds]
wens_ has joined #linux-sunxi
itdanihe1 has joined #linux-sunxi
dlan has joined #linux-sunxi
woprr_ has joined #linux-sunxi
spv__ has joined #linux-sunxi
Wizzup_ has joined #linux-sunxi
TheSeven has quit [*.net *.split]
egbert_ has quit [*.net *.split]
MadSpark has quit [*.net *.split]
focus has quit [*.net *.split]
jelly-home has quit [*.net *.split]
slapin has quit [*.net *.split]
torbenh3 has quit [*.net *.split]
tkoskine has quit [*.net *.split]
oliv3r has quit [*.net *.split]
mnemoc has quit [*.net *.split]
torbenh3 has joined #linux-sunxi
MadSpark has joined #linux-sunxi
tkoskine has joined #linux-sunxi
Eyal_ has joined #linux-sunxi
bamvor has joined #linux-sunxi
woprr_ has quit [Client Quit]
TheSeven has joined #linux-sunxi
woprr_ has joined #linux-sunxi
hurtigbuffer has joined #linux-sunxi
focus has joined #linux-sunxi
book` has quit [Ping timeout: 246 seconds]
book` has joined #linux-sunxi
Seppoz_nowork has joined #linux-sunxi
eyal has quit [*.net *.split]
bamvor_ has quit [*.net *.split]
wens has quit [*.net *.split]
ojn has quit [*.net *.split]
spv_ has quit [*.net *.split]
Seppoz has quit [*.net *.split]
kriegerod has quit [*.net *.split]
woprr has quit [*.net *.split]
Wizzup has quit [*.net *.split]
tm512 has quit [*.net *.split]
itdaniher has quit [*.net *.split]
VargaD has joined #linux-sunxi
slapin_ has joined #linux-sunxi
oliv3r has joined #linux-sunxi
egbert has joined #linux-sunxi
hurtigbuffer is now known as jelly-home
kriegerod has joined #linux-sunxi
mnemoc has joined #linux-sunxi
ojn has joined #linux-sunxi
<wens_>
Turl: setting 2 "default"s in pinctrl-names works
dlan_ has joined #linux-sunxi
kriegerod has quit [*.net *.split]
Eyal_ has quit [*.net *.split]
torbenh3 has quit [*.net *.split]
dlan has quit [*.net *.split]
bfree has quit [*.net *.split]
ssvb has quit [*.net *.split]
kriegerod has joined #linux-sunxi
Eyal_ has joined #linux-sunxi
ssvb has joined #linux-sunxi
bfree has joined #linux-sunxi
torbenh3 has joined #linux-sunxi
Phoenix_ has joined #linux-sunxi
t3st3r has quit [Remote host closed the connection]
t3st3r has joined #linux-sunxi
t3st3r has quit [Remote host closed the connection]
phoenix__ has joined #linux-sunxi
t3st3r has joined #linux-sunxi
Phoenix_ has quit [Ping timeout: 272 seconds]
wens_ is now known as wens
<libv>
aha, libnand parameters found.
<gzamboni>
hi, i think we should put content table in the front page of the wiki with all the kernels versions and its states and use (with the branchs of the sunxi's forks and the mainline) what do you all think ?
<libv>
gzamboni: why?
<libv>
gzamboni: should this not live in linux-sunxi.org/Linux_Kernel ?
<gzamboni>
libv, is it linked in the mainpage ? i cant see the link to this page
<libv>
it indeed is not referenced from the front page
<libv>
gzamboni: stick Category:Tutorial at the end of that page
<gzamboni>
k
<gzamboni>
shouldnt it be in the software first table ?
<gzamboni>
it isnt really a tuto
<libv>
gzamboni: it is referenced from a lot of places, or at least, it should be
<libv>
gzamboni: the thing is, if you change this, you also need to change the status of the u-boot page
mrnuke has quit [Remote host closed the connection]
<libv>
and the people who actually work on it know where it is
<gzamboni>
gotta go let the kids at school, brb
<gzamboni>
libv ok
pacopad has joined #linux-sunxi
Super-noob has joined #linux-sunxi
Super-noob has quit [Ping timeout: 264 seconds]
phoenix__ has quit [Ping timeout: 272 seconds]
mrnuke is now known as CaptainObvious
<oliv3r>
gooood morning
<wens>
morning
Super-noob has joined #linux-sunxi
woprr_ has quit [Quit: Ex-Chat]
CaptainObvious has quit [Quit: Coyote finally caught me]
mrnuke has joined #linux-sunxi
mrnuke has quit [Client Quit]
mrnuke has joined #linux-sunxi
n01 has joined #linux-sunxi
Super-noob has quit [Ping timeout: 264 seconds]
Super-noob has joined #linux-sunxi
mrnuke has quit [Quit: Coyote finally caught me]
tm512 has joined #linux-sunxi
mrnuke has joined #linux-sunxi
ynezz_ is now known as ynezz
Phoenix_ has joined #linux-sunxi
<Phoenix_>
Hi, I'm having a problem maybe someone can help with. I suspect one of the drivers is causing a null pointer error when populating /dev during boot. Is there a way I can work out which driver is the problem? http://pastebin.com/fMJYhjcU
arete74 has joined #linux-sunxi
Quarx has joined #linux-sunxi
bbrezillon has joined #linux-sunxi
notmart has joined #linux-sunxi
tomboy64 has joined #linux-sunxi
HeHoPMaJIeH has joined #linux-sunxi
HeHoPMaJIeH has quit [Client Quit]
<oliv3r>
vector80: i think we do have som eencoding examples, but they are marginal at best; i think h264 encoding is slowly being looked at being RE-ed
HeHoPMaJIeH has joined #linux-sunxi
woprr has joined #linux-sunxi
parabyte has joined #linux-sunxi
parabyte has quit [Client Quit]
parabyte has joined #linux-sunxi
FR^2 has joined #linux-sunxi
xeros has quit [Ping timeout: 245 seconds]
xeros has joined #linux-sunxi
hipboi has joined #linux-sunxi
rellla has joined #linux-sunxi
parabyte has quit [Quit: Leaving]
ssvb has quit [Ping timeout: 265 seconds]
xeros has quit [Ping timeout: 264 seconds]
ssvb has joined #linux-sunxi
pacopad has quit [Ping timeout: 248 seconds]
<n01>
argh, no comments on irqchip patch yet?
xeros has joined #linux-sunxi
Eyal_ has quit [Read error: Operation timed out]
maz_ has joined #linux-sunxi
wolfy has joined #linux-sunxi
wolfy has left #linux-sunxi [#linux-sunxi]
eyal has joined #linux-sunxi
wingrime has joined #linux-sunxi
deasy has joined #linux-sunxi
pacopad has joined #linux-sunxi
popolon has joined #linux-sunxi
popolon has joined #linux-sunxi
bsdfox has quit [Ping timeout: 245 seconds]
maz_ has quit [Ping timeout: 264 seconds]
<wens>
n01: you fixed the deadlock?
<n01>
wens: yep
<wens>
what was it?
<n01>
didn't ack the GIC -> no more interrupts :)
<wens>
ah
<n01>
solved with chained_irq_enter and chained_irq_exit
<plaes>
\o/
bsdfox has joined #linux-sunxi
<plaes>
n01: shouldn't the dt doc commit be together with the code?
<n01>
plaes: in v2 is in patch 3/3
<plaes>
yeah, I see, but shouldn't they be in the same commit?
<wens>
some say it should be split
<n01>
honestly I'm not sure
<plaes>
ok, I remember someone complaining about the split :)
maz_ has joined #linux-sunxi
<mripard>
plaes: usually, the DT maintainers prefer to have the bindings documentation separate
<plaes>
ok, thanks
<oliv3r>
it makes sense it being together
<oliv3r>
you add a new feature, that includes documentation to document said feature
<mripard>
except that you don't document a feature
<mripard>
you document a binding
torbenh3 has quit [Read error: Operation timed out]
<mripard>
and that you will use that binding later on to add your new feature :)
paulk-collins has joined #linux-sunxi
<n01>
mripard: btw, I'm waiting a bit before submitting the MFD since I want to be sure that the skeleton remains the same after adding the regulators support
Black_Horseman has joined #linux-sunxi
<oliv3r>
mripard: tomato tomatoh!
<oliv3r>
mripard: so you are adding a binding and documenting its usage!
<oliv3r>
imo they go together like bread and butter :)
<mripard>
yeah, but it doesn't mean that you will actually use it.
<mripard>
take the timers for example, there is 4(?) interrupts for the timers in the hardware
<mripard>
you register the 4 of them in the DT
<mripard>
so that it accurately describe the hardware
<mripard>
yet, the timer driver might use only one
Black_Horseman has quit [Quit: Zwi se logou mou!!!]
<hramrach>
and looking at the diff there was not more useful info before so just leave it :p
<libv>
i just mailed him through the wiki
<libv>
referred him to the new_device_howto
<libv>
now, back to hacking in support for my nand
<libv>
what's up with libnand btw?
<libv>
how did we ever get the code for that original? tom?
<libv>
originally
FreezingCold has quit [Ping timeout: 264 seconds]
* nove
would like to see as demo a spinning cube which the textures are from a video,, all open source of course, but the time is now short
<libv>
nove: a fully working glmark2-es2 is a much more important milestone
<nove>
libv: just a idea for a demo that oliv3r could show, to give a wow factor, i am aware that things are not ready and this is likely not possible
<libv>
it shouldn't be overly hard to do with limare and a yuv texture, but it will not be possible in the few weeks left
bbrezillon has quit [Ping timeout: 260 seconds]
<wingrime>
nove: pin
<wingrime>
g
<nove>
wingrime: pon
<oliv3r>
libv: originally it just got dropped; later versions they closed it; after mailing them, they sent the code
<oliv3r>
so we do have an updated libnand; mnemoc knows where it is
<wingrime>
nove: actualy, ISP buffer , that Result buffer , or source , or Source + Result of scale
<wingrime>
??
<oliv3r>
nove: if you can manage to hack cedar playback into something would be cool but i think we are stuck with the standard xorg stuff i assume
<nove>
wingrime: source only
<wingrime>
nove: ok, but where result is?
<wingrime>
nove: thats not possible ISP store resize result to sram
<nove>
wingrime: there is no result buffer, buffer -> ISP -> AVC
<wingrime>
nove: and more why WB_THUMB
<wingrime>
nove: ^^ -- WriteBack
<wingrime>
nove: no, it still not possible without some temporary buffer
<nove>
wingrime: each macroblock at a time, maybe
FreezingCold has joined #linux-sunxi
<wingrime>
nove: still looks strange for me
<wingrime>
nove: normaly only one action can be done automaticly -- I think AVC can only poke ISP TRIG, and recive result to AVC buffer
<wingrime>
nove: yes, buffer can be on AVC side, but it sure must exsist
bbrezillon has joined #linux-sunxi
<nove>
wingrime: that buffer is no where in the traces
<wingrime>
nove: AVC_VLE_ADDR ??
<wingrime>
nove: I think it need to be alocated, but not filled
<nove>
wingrime: in my test jpeg encoder (no binaries), source luma and chroma, and the output of the encoded bitstream, only this buffers
<nove>
wingrime: only this 3 buffers*
<wingrime>
nove: 3?
FreezingCold has quit [Ping timeout: 260 seconds]
<wingrime>
oh
<wingrime>
nove: but AW claim that we can also alpha blend....
<wingrime>
and nove and how about non integer
<wingrime>
resize
<wingrime>
nove: can you check source buffers integrity???
<nove>
wingrime: that 0xb8c register, i don't know, it is initial set a 04000000, like the start of DRAM, but if i remenber correctly the lower bits are incremented in the same way as MACC_AVC_VLE_LENGTH
<nove>
wingrime: yes, i can do that in a while
<wingrime>
nove: also, check, is there something new after source buffers ...
<wingrime>
nove: and ISP's sram...
eebrah|mobile has joined #linux-sunxi
<wingrime>
nove: looks like ISP have others input/output registers
jinzo has joined #linux-sunxi
jinzo has quit [Changing host]
jinzo has joined #linux-sunxi
<nove>
wingrime: non integer, there that table that is writen to sram, with huge scale, one can see that the pixels don't have smooth gradients, but have are pixeled in 16/32 steps
<oliv3r>
libv: how important is it to keep the existing installation prestine?
<oliv3r>
libv: you could, in thoery, boot the original OS, backup all data to 'nfs/usb/sd', format with our older libnand, copy files back
<oliv3r>
it's a a pain, but the format hasn't changed hugely
Gerwin_J has joined #linux-sunxi
<oliv3r>
boot0/1 does seem to work with newer versions too
hani has joined #linux-sunxi
<Turl>
morning
hansg has joined #linux-sunxi
<Turl>
oliv3r: lichee should, I believe it has boota for that
FreezingCold has joined #linux-sunxi
hani has quit [Ping timeout: 252 seconds]
<Turl>
libv: I believe lkcl got the nand code
<libv>
aha, lkcl knows how to get it then
hansg has quit [Quit: Leaving]
<libv>
it's going to take me a day to figure out this last flags i need for my previously unknown chip
<Turl>
libv: what flags?
<Turl>
what chip? :)
<oliv3r>
Turl: hmm, i did notice my u-boot was about 100k smaller then the nand version; but my tablet just stops at 'starting kernel'
<Turl>
oliv3r: using boota?
<Turl>
oliv3r: remember also different machid
<libv>
Turl: optional flags for my unknown nand chip
<libv>
everything else was immediately readable
<oliv3r>
Turl: original kernel, i haven't changed the tablet; so everything is stock; i only built a u-boot that does uSD uart, and replaced that + script.bin with uart support
<libv>
oliv3r: this stopping at starting kernel...
eebrah|mobile has quit [Ping timeout: 248 seconds]
<oliv3r>
libv: thanks; i'll use that tonight for my tests as that's 'guaranteed to work'
<libv>
well, at least give it a whirl once to verify that it isn't those bits
<oliv3r>
aye indeed
<oliv3r>
and allows me to test felboot!
bbrezillon has quit [Ping timeout: 272 seconds]
<nove>
wingrime: with my simples test, there is no new change in content of the buffers, tested all dram, and also reran without cleaning/writing frame data (using the last buffer contents)
tomboy64 has quit [Quit: And remember, aal is well.]
bbrezillon has joined #linux-sunxi
pacopad_ has joined #linux-sunxi
enrico__ has joined #linux-sunxi
pacopad has quit [Ping timeout: 251 seconds]
pacopad_ is now known as pacopad
pacopad_ has joined #linux-sunxi
pacopad has quit [Ping timeout: 260 seconds]
pacopad_ is now known as pacopad
kivutar has joined #linux-sunxi
<itdanihe1>
What's the libnative alternative to `do task::spawn_sched(task::SingleThreaded) {
itdanihe1 is now known as itdaniher
FDCX__ has quit [Ping timeout: 245 seconds]
rellla has quit [Ping timeout: 260 seconds]
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<itdaniher>
whoops, wrong channel :)
rellla has joined #linux-sunxi
HeHoPMaJIeH has quit [Remote host closed the connection]
<nove>
jemk: if you didn't notice, the viewer register fields were updated
wolfy has left #linux-sunxi [#linux-sunxi]
<jemk>
nove: yes, I noticed. I think i will add much more to wiki tomorrow or so from h264 encoding, if it works then.
n01 has joined #linux-sunxi
ccaione has joined #linux-sunxi
n01 has quit [Client Quit]
<bsdfox>
yay found uart0 finally. will document shortly
ccaione has quit [Quit: Bye]
FreezingCold has joined #linux-sunxi
n01 has joined #linux-sunxi
FR^2 has quit [*.net *.split]
FR^2 has joined #linux-sunxi
<libv>
techn_: yeah, but in the meantime i figured out what was broken with my sd card
<libv>
bsdfox: starts to look good
<libv>
bsdfox: are you sure you want to keep that name though?\
<libv>
bsdfox: ah yes, uart would be those two pads on the backside
<wingrime>
jemk: ping
<wingrime>
jemk: same question for you
ccaione has joined #linux-sunxi
<jemk>
wingrime: pong
n01 has quit [Quit: leaving]
<wingrime>
jemk: are ISP have output or temporary buffer
<wingrime>
?
gzamboni has quit [Remote host closed the connection]
gzamboni has joined #linux-sunxi
<jemk>
wingrime: maybe it has, but it is not needed/used for encoding. at least I only set input buffer
<jemk>
wingrime: maybe for other functions, there must be some thumbnail function
<wingrime>
jemk: but VLE have input registers
<wingrime>
jemk: are you alocated it?
<wingrime>
jemk: VLE input buffer?
<jemk>
wingrime: what do you mean?
<wingrime>
jemk: I mean, we have MACC_AVC_VLE_ADDR, that result buffer?
<jemk>
wingrime: for h264 enc i have till now: input luma/chroma in isp regs, output bitstream and temp buffer for reference pictures in avc engine
<jemk>
wingrime: that's the output for the encoded bitstream
<wingrime>
jemk: I still think ISP shuld have output buffer or so
<wingrime>
jemk: it must be workable standalone
<wingrime>
jemk: is is possible enable ISP without AVC be enabled ?
<jemk>
wingrime: possible, but i only know the h264 enc related things
<wingrime>
jemk: as I know, it can do alpablending
<jemk>
wingrime: i think alpha blending is in general regs 0-0x100
<wingrime>
jemk: I mean, MACC_VE_CTRL_ENGINE 0xa = ISP, enables only ISP register set ?
<wingrime>
jemk: I think alpha blending in ISP, and I think iSP can resize in standalone configuration,
<wingrime>
jemk: we still know very little set of ISP registers
<jemk>
wingrime: there are many regs we don't know yet, that's the curse of reverse engineering
<wingrime>
jemk: also, there intersting news about a23 kernel, it have more one blob - libisp , but I have no idea , is that same cedar's isp
<wingrime>
jemk: also, new a23 kernel define ISP IRQ register
<jemk>
btw, is this a random failure or does our rtc (cubietruck) have problems with 2014? mine is at 1983 since 1.1.
<ccaione>
new nick on znc
<wingrime>
mripard: ping
netlynx has quit [Quit: Leaving]
<wingrime>
mripard: sunxi_timer.c in a23 kernel with your copyright
<wingrime>
sun4i too
<bsdfox>
libv, the trio stealth pro name?
Akaizen has quit [Remote host closed the connection]
<wingrime>
oliv3r: ping
<bsdfox>
hmm my uboot has no delay on this device. I mounted /dev/block/nanda and edited boot.ini (only config file there that looked relevant) and set timeout = 9 (instead of -1) but it still doesn't pause
<bsdfox>
any advice? I'm wanting to figure out why it won't boot from microsd
rellla has joined #linux-sunxi
Akaizen has joined #linux-sunxi
<jemk>
yeah, "playing" 4k video on cubietruck. cedar does it perfectly well, only disp isn't fast enough to downscale
<jemk>
wingrime: ^
<Akaizen>
jemk: thats awesome - what kernel and modules?
<wingrime>
jemk: are you fixed something
<wingrime>
?
<jemk>
Akaizen: sunxi-3.4 with fixed vdpau
<jemk>
wingrime: yep
<wingrime>
jemk: last time I tryed play 4k video I get strage glitches - half of screen
<wingrime>
jemk: and, are you impoverd input buffer size 1Mb not good when you have ~150MB for cedar
<wingrime>
jemk: what are you fixed?
<wingrime>
jemk: and, you can add vdpau env parameter for downscale or something for make engine-internal-downscale
<jemk>
wingrime: patch on its way, but now it crashed...
<wingrime>
jemk: are you tested - tears of still in 4k??
<jemk>
wingrime: not yet, some short 4k sample from somewhere. but after i find why it crashed i will try
<wingrime>
jemk: you can try use internal engine downscaler
<jemk>
wingrime: one after another
<wingrime>
jemk: If cedar can 4k normaly if would cool news for luke and Tsvetan
<jemk>
wingrime: I think it is a little bit laggy even with 320MHz, but not sure if thats cedar or disp
<wingrime>
jemk: you still have 1MB for buffer?
<wingrime>
jemk: if it less than 1 frame if will be indeed laggy
<jemk>
wingrime: yes, but that can't be a problem. with vdpau this buffer always contains only one frame as you could see in decoder.c
<jemk>
wingrime: and no, i can't change it as vdpau only provides one frame and wants the decoded picture then
<wingrime>
jemk: that's can kill much performance
<wingrime>
jemk: when it not zero-copy
<jemk>
wingrime: input data is only kilobytes, so that shouldn't be such a big problem
<wingrime>
jemk: kilobytes when P/B frame
<wingrime>
jemk: I frame should be big anought
<wingrime>
jemk: also, read in bulk are much better
<jemk>
wingrime: we can't change that, its vdpau limitation
kivutar has quit [Ping timeout: 252 seconds]
wolfy has joined #linux-sunxi
<nove>
is true, jpeg encoder only does max 4080x4080
<wingrime>
nove: is limit are artifical?
FDCX_ has joined #linux-sunxi
wolfy has left #linux-sunxi [#linux-sunxi]
<nove>
wingrime: MACC_ISP_SCALER_SIZE only uses 1 byte for each size,, but that is when using the scaler,
<nove>
wingrime: without the scaler the size is 2 bytes, but bigger sizes only repeats picture, at least in the vertical
<jemk>
wingrime: ok, 4k will always be laggy a lot. It needs between 60ms and 85ms per frame, so ~11fps in worst case
<mripard>
wingrime: oh? nice
<mripard>
I haven't look in depth at the A23 kernel yet
<wingrime>
mripard: yes, aw used some you code
<mripard>
except for the dmaengine and pinctrl drivers you pointed out
<wingrime>
mripard: also I can say they replaced ar100 with arisc
<wingrime>
mripard: and ar100 can play audio from sram
<wingrime>
mripard: thats explais their demo
<wingrime>
mripard: but I not saw that can decode mp3
<wingrime>
err arisc,not ar100
<oliv3r>
yay, i got my 4E USB voltage/amperage meter
<oliv3r>
i can measure power draw over USB!
<wingrime>
oliv3r: ^^^^^^^^^^^^^^
<oliv3r>
wingrime: sup
<wingrime>
oliv3r: ar100 now outdated, arisc are used
t3st3r has quit [Remote host closed the connection]
<oliv3r>
bsdfox: extract the uenv partition (nandb my guess), just dd the entire partition (but not more then that partition,s hould be the smallest one)
<wingrime>
oliv3r: arisc can play music from sram when standby
<oliv3r>
bsdfox: take a hexeditor and editor and edit the boot delay var, write it back :)
<oliv3r>
wingrime: on a31?
t3st3r has joined #linux-sunxi
<wingrime>
oliv3r: i oliv3r a23
<wingrime>
err
<oliv3r>
wingrime: and ar100 was a dedicated core, so what is arisc? ( i will read log back later)
<bsdfox>
oliv3r, doesn't the A10 try to boot from SD every time?
<oliv3r>
ohh a23! that's the a?100 chip?
notmart_ has quit [Quit: notmart terminated!]
<wingrime>
oliv3r: more one dedicated core
<oliv3r>
bsdfox: the BROM will, without exception, but from MMC0, NAND, MMC2, SPI
<wingrime>
oliv3r: they added dvfs , sound play, axp control to it
<oliv3r>
bsdfox: IF however, you press the fell button OR all of the above have an invalid signature, it will drop to FEL mode
<oliv3r>
wingrime: so a23 has ar100 too? so it's like a31
<oliv3r>
so we should start supporting a31 better :)
<wingrime>
oliv3r: nonono
<wingrime>
oliv3r: I mean, they changed dedicated core from ar100 to arisc in a23
<wingrime>
I mean thats different core
<oliv3r>
wingrime: yeah btu still a seperated extra core
<oliv3r>
well we know there is an extra chip that we don't know what it does
<wingrime>
oliv3r: dvfs + axp control + play sound from sram
<oliv3r>
a companion chip, could be risk, could be audio chip
<wingrime>
oliv3r: a23 kernel have code for it
<wingrime>
oliv3r: and test code
<oliv3r>
but that's really nice to have, we can sleep all 4 cores and be in standby, and still access SRAM and do stuff
<oliv3r>
wingrime: so you gonna write driver for that? :D
<wingrime>
oliv3r: no HW
<wingrime>
oliv3r: driver already there !!
<wingrime>
oliv3r: or you mean mainline?
<oliv3r>
yeah mainline driver :p
<oliv3r>
well it's 2 things isn't it
timothybean has joined #linux-sunxi
<oliv3r>
it's a real 'firmware'
<oliv3r>
so you need 2 things
<timothybean>
Hi everyone
<wingrime>
oliv3r: firmware are difficult
<oliv3r>
a driver to access/talk to the arisc; and you need to upload firmware to it
<timothybean>
Can anyone help me with a few questions regarding the A20?
<wingrime>
oliv3r: I may be hardcoded
<wingrime>
oliv3r: I mean PROM firmware
<wingrime>
err
<wingrime>
oliv3r: like bootloader it can be not flashable
<mripard>
wingrime: and so, it's still a weird openrisc core? or is it a different architecture?
<mripard>
oliv3r: on the A31, it's not that difficult to do
<wingrime>
mripard: yes, looks like
<oliv3r>
mripard: i dunno if you followed this from the past
<wingrime>
mripard: they used HW looks and msgbox
<wingrime>
*locks
<mripard>
wingrime: yep
<wingrime>
mripard: question is firmware
<mripard>
plus remoteproc to upload the firmware if needs be, and we're done
<oliv3r>
mripard: but a23 looks to be a lot more like a31, then the others; i think it is a 'true' hybrid
<mripard>
on the A31, the firmware was built in the kernel image iirc
<oliv3r>
mripard: also, we got a23 sdk, with newer kernel sources; they started to unify their sunxi arch
<mripard>
and sent to the ar100
<oliv3r>
mripard: isn't it 'nicer' to have a way to upload the firmware? so you can upload whatever you want
<wingrime>
oliv3r: nice way --- change it on fly
<mripard>
oliv3r: yep, remoteproc does that.
<wingrime>
mripard: they used shared memory?
<bsdfox>
oliv3r, yay that worked. thanks
<wingrime>
or SRAM?
<mripard>
wingrime: I don't remember :(
<wingrime>
mripard: they used CMA
<bsdfox>
sunxi#mmcinfo
<bsdfox>
No MMC device available
<mripard>
wingrime: for the A23?
<bsdfox>
I suppose that's a problem why my sd card doesn't boot?
<wingrime>
mripard: I see something close
<oliv3r>
mripard: never heard of remote proc :p
<oliv3r>
i would not be supprised if they used SRAM
<mripard>
oliv3r: it's the framework to handle the coprocessors :)
<oliv3r>
bsdfox: wrong u-boot?
<wingrime>
mripard: they start using CMA framework for VE
<oliv3r>
mripard: remote-processor, makes sense :p
<oliv3r>
mripard: i suspect co-processor is not like the 387 :p
<bsdfox>
oliv3r, uboot should be read from the sd card :P
<bsdfox>
it's the same SD card image that boots on my mele with the exception of a different uboot
<oliv3r>
but i think except for dedicated IP (emac etc) the kernel doesn't use SRAM at all does it?
<mripard>
it does
<mripard>
for the PSCI frimware
<oliv3r>
bsdfox: well your u-boot is failing
<wingrime>
mripard: they used new so-core for audio play, so I think they should add more memory to SRAM
<mripard>
but it uses the secure sram
<oliv3r>
mripard: ah that's new stuffs; i don't know what psci is (i saw the smp patches with that name though)
timothybean has left #linux-sunxi [#linux-sunxi]
<wingrime>
mripard: but I still can't found new sram size
timothybean has joined #linux-sunxi
<oliv3r>
ah it abuses turstzone ram; or is it actually part of trustzone?
<oliv3r>
i really should buy an a23
<oliv3r>
if only i had moneys
<nove>
timothybean: we can only know if we can answer if you ask the question
<mripard>
oliv3r: it doesn't abuse anything :)
<mripard>
it just puts the firmware that will handle the PSCI calls in that area
<timothybean>
Can anyone tell me how many simultaneous videos (H.264) could be decoded on a A20?
<timothybean>
sorry, my keyboard was messed up
<bsdfox>
oliv3r, http://pastebin.com/TrNBxyfv where in the process does it try to boot from SD? is it before BOOT0 or after? is the CRC error something I should be looking at?
<ssvb>
wingrime: Allwinner started using CMA in their new SDK?
Gerwin_J has joined #linux-sunxi
<wingrime>
ssvb: yes
<nove>
timothybean: the userland driver and the kernel driver is still not ready to do multi decoding
<timothybean>
Also, is Gstreamer available with hardware acceleration?
<timothybean>
OK, so just one at a time...
<wingrime>
ssvb: they started to use some mripard code , also , nice dmaengine driver
paulk-collins has quit [Quit: Ex-Chat]
<wingrime>
ssvb: they stat clean code for all
<ssvb>
wingrime: CMA makes a lot of sense :) just wonder which patchset they have used and how it compares to what we have in stage/sunix-3.4
<nove>
timothybean: but that answer depends of course the resolution of the videos
<ssvb>
wingrime: which kernel version are they using in this SDK?
<nove>
timothybean: no gstreamer, no ffmpeg, is still work to be done
<wingrime>
ssvb: 3.4
<timothybean>
OK, I had just been looking into it, but hadn't seen anything yet... seems a little information right now that I could find.
<ssvb>
wingrime: ok, most likely it's the same backport from LTSI then
<timothybean>
Is all this effort because Allwinner hasn't released information so it has to be reverse engineered?
<oliv3r>
mripard: well isn't sram B officially soley for trustzone
<timothybean>
I would like to know about decoding D1 and 720p resolutions
<oliv3r>
mripard: so if you use it for something else, you technically can't use trustzone anymore
<oliv3r>
bsdfox: boot0 is from nand
<oliv3r>
bsdfox: so your not booting form MMC at all
<timothybean>
Also, I saw that A20 has 4 CVBS out and inputs... are they working?
<oliv3r>
did you dd if=u-boot-with-spl.bin of=/dev/yoursd seek=8 bs=1024?
<oliv3r>
timothybean: hardware yes, drivers, so so
<oliv3r>
timothybean: input i'm not sure if we allready have a driver
<oliv3r>
timothybean: all I know that cbs is used for vga (3 pins, R G B), composite and SVIDEO (2 pins) not sure if you can use them simultaniously, but I don't see why not
<oliv3r>
timothybean: but i doubt our driver does that; as we don't really have
<oliv3r>
wingrime: so it needs to store 32k samples
<oliv3r>
play that, resume frmo standby, load 32k new samples, standby again
<wingrime>
oliv3r: capture + play bufferes
<oliv3r>
if they support 44khz, that's 32k * 44khz (x2 for stero)
<timothybean>
Data sheet says 4 CVBS outputs... but could the driver be updated? But, I suppose it doesn't matter though if you can't decode multiple videos?
<oliv3r>
timothybean: we only have original aw driver; which is horrible
<timothybean>
I would even be willing to write a Gstreamer plugin for A20 to do HW decoding if needed.
<oliv3r>
timothybean: we dont' have a decoder driver yet
<mripard>
oliv3r: if you try to access this sram from the !secure world, you'll only get 0s
<oliv3r>
timothybean: we have a Proof of concept, that is improving, we can decode jepg, mpeg and h264, but it's far far from done
<oliv3r>
timothybean: but it IS a vdpau thing :)
<wingrime>
mripard: same for you
<timothybean>
right, and you don't know how to talk to it?
<oliv3r>
mripard: if you activate trustzone, yeah but then it won't be usefull for pcsi
<wingrime>
mripard: they use strange memory offset for audio sram
<mripard>
oliv3r: so you can't abuse it from the non secure world
<oliv3r>
timothybean: we RE-ed most of it
<timothybean>
I mean, why would Allwinner not provide support for this? How would anyone make anything for it without drivers
<oliv3r>
mripard: well if you disable trustzone, it becomes fully useable and mappable, i thought that's what the pcsi thing did
<wingrime>
timothybean: thats we do now
<oliv3r>
timothybean: there's the cedarX closed source blobs; which (besides being a GPL violation) are an abomination
Akaizen has quit [Remote host closed the connection]
<mripard>
oliv3r: no, the PSCI firmware runs in secure world, not the kernel
<timothybean>
Well, it seems discouraging from using this in a new product ...
<oliv3r>
timothybean: all the luck to you trying to make them work :) (it IS possible)
<timothybean>
Is there no way to get it from Allwinner?
<oliv3r>
mripard: ahh, so psci is a trustzone thing actually; very intersting
<oliv3r>
timothybean: we have the blobs, they are just horrible crap
<wingrime>
oliv3r: how much time tablet will play in standby?
<timothybean>
why are they crap, what do you mean by this?
<timothybean>
they don't work right?
<oliv3r>
timothybean: what they did was, grab free code (and ffmpeg code), mostly first reference codecs etc (literally) ripped out pieces to interface with the hardware, and made an android blob out of it
<oliv3r>
timothybean: they don't decode everything, crashy, under android they work reasonbly okay-ish
<oliv3r>
timothybean: the linux binaries are horrible
<wingrime>
timothybean: they targets cheap android tablets, in massssss production
Wizzup_ is now known as Wizzup
<ssvb>
oliv3r: not sure if you noticed that I have pinged you earlier
<ssvb>
oliv3r: "<ssvb> oliv3r: which of the sunxi devices are you going to use for your FOSDEM demonstration?"
<oliv3r>
timothybean: lets say, it was written/hacked together by 1 or 2 chinese youngsters, that aren't awesome devs :)
<wingrime>
timothybean: they don't care about some enterprices
Wizzup is now known as Guest3182
<oliv3r>
ssvb: i don't know yet! (Will you be there) I still have to backread; wingrime distracted me :p
<oliv3r>
ssvb: i'm thinking of showing a tablet doing some video work maybe, but likly a cubietruck
<ssvb>
oliv3r: yes, I will be there
<oliv3r>
ssvb: because it has VGA output
<timothybean>
understood, so what would it take to get them working right? or at least reasonably well?
<oliv3r>
ssvb: haven't seen you reply to the saturday dinner mail :p
Guest3182 is now known as Wizzup
<oliv3r>
ssvb: the cubietruck can be hooked up to the beamer via vga
<timothybean>
how are you able to figure out what is happening from the blobs?
<ssvb>
oliv3r: hmm, hansg asked to send private e-mails to him to apply :)
<oliv3r>
timothybean: that's mostly wingrime, nove's and jemk's work; i was only very initally working on that; but lots of RE, there's some tracing software etc
<timothybean>
I mean, is it a task that is too big and should not be attempted?
<ssvb>
oliv3r: anyway, cubietruck is a good choice
<oliv3r>
ssvb: ohh, i think most CCed the ML :p
<oliv3r>
ssvb: but i'm very open to any suggestions/advise
<oliv3r>
wingrime: PALY really has to be PLAY :p
<wingrime>
oliv3r: that's AW's
<oliv3r>
wingrime: looks like they have several buffers; 2, 4, 8 16 and 32k all simultaniously
<ssvb>
oliv3r: I just thought that you wanted to have a demo with lime, because you have its picture on the fosdem website :)
<oliv3r>
ssvb: i would love to
<oliv3r>
but i don't have one :p (and they don't have vga do they
<oliv3r>
If tsvetsan brings one with vga; i may steal one from him and do it on that
<oliv3r>
i'll use an SD card anyway, so it shouldn't matter
<nove>
timothybean: with binaries blobs you are trapped by what allwinner releases, the cedar hardware is been upgrade each new SOC, and the binaries apis also changes
<oliv3r>
speaking of; wingrime what did you use for your cedarX stuff? should run on hansg's fedora image right?
<oliv3r>
I actually have to get back up to speed with nove, jemk and wingrime's work here :p
<ssvb>
oliv3r: well, I tried to steal the 16-bit lime dram config and use it on a cubieboard to run some benchmarks
<oliv3r>
so expect some trolling from me :p
<oliv3r>
and ssvb's xorg driver
<oliv3r>
ssvb: and?
<nove>
timothybean: a safe bet is the open source library, but the is still alot of work to be done
<ssvb>
oliv3r: it runs very slow for anything graphics related, the framerate is roughly halved in mali benchmarks :(
<ssvb>
oliv3r: but I just wonder if using cubieboard in this configuration is a valid lime simulation
<ssvb>
oliv3r: it would be nice to try the real thing
<jemk>
anyone familiar with the clock system, I think ve clock is broken in sunxi-3.4...
<ssvb>
jemk: hmm, what's wrong with it?
<ssvb>
jemk: patrick tried to get it "right", but maybe something went wrong
<jemk>
ssvb: always same speed, somewhere around 150mhz if one can trust clock_gettime
<oliv3r>
ssvb: reasonably valid
RzR has joined #linux-sunxi
<oliv3r>
ssvb: since we know the lime is using a 16bit memory bus, right?
<oliv3r>
ssvb: so performance should be roughly halved
<oliv3r>
i'm really concerned about 'xbmc' performance
<oliv3r>
one of the reasons the pi is so popular is, hacker potential + raspbmc
<ssvb>
jemk: even after using the clock speed setup ioctl? on what hardware?
<oliv3r>
so if xbmc runs really horrible on the LIMA (even as 720p for desktop stuff, full hd for video) then it won't succeed
<jemk>
ssvb: yes, after ioctl on A20 (cubietruck).
<ssvb>
oliv3r: maybe you have noticed in the mailing list that I'm kinda trying to make a demo sdcard image myself :)
<oliv3r>
ssvb: i know :)
<oliv3r>
ssvb: not sure if I have replied yet; but i 100% agree with you
<oliv3r>
ssvb: and i dunno about you; but i _really_ want the LIME to succeed
<oliv3r>
but the v0 with 16 bit memory bus might be troublesome
<oliv3r>
i don't think there's DRAM chips with 32bit bus are there?
<ssvb>
oliv3r: smplayer + vdpau works quite nicely, we probably don't need to show xbmc (which uses proprietary blob and has a lot of quirks)
<oliv3r>
ssvb: also, my cubietruck has an old iPhone battery; making it battery operateable :p
<oliv3r>
ssvb: for the demo; no xbmc; maybe next year ontop of lima :p
<ssvb>
oliv3r: we should try to convince the xbmc developers to make use of the open source hardware decoder
<oliv3r>
ssvb: but i ment generally; lime as pi replacement
<RzR>
oliv3r, talking about fosdem demos ?
<oliv3r>
RzR: hmm?
<RzR>
sorry
<oliv3r>
ssvb: you mean cederus vdpau?
<ssvb>
oliv3r: as I said, smplayer plays hd movies and it has a nice gui :)
<RzR>
wrong room :)
<oliv3r>
ssvb: but that's so far from done :(
<oliv3r>
RzR: yeah i was; but hmmm?
<RzR>
thought I was into #olimex
<oliv3r>
RzR: close enough
<RzR>
yes I plan to attend those tracks
<oliv3r>
i shouldn't give away TOO much :p
<ssvb>
oliv3r: why is it far from done? it just needs some tweaks for sure, but should be quite usable
<RzR>
I was into test leon's tizen image adapted to a20
<RzR>
testing
<oliv3r>
ssvb: i thought it was all still alpha PoC code; and a full rewrite would be required anyway; but i admit; i haven't kept up to date
<mripard>
oliv3r: if you want to learn how to work with remoteproc
<ssvb>
jemk: then maybe something is indeed wrong, try to add some debug prints around the code which is configuring the pll divisor
<oliv3r>
mripard: i'll def. watch this at work tomorrow
kivutar has joined #linux-sunxi
<mripard>
mdp was using it to drive a 6502
<mripard>
(and his talk was awesome.)
<ssvb>
oliv3r: a major cedar rewrite is needed because it is using mmio and happily passes non-sanitized physical addresses around
<ssvb>
oliv3r: but other than this, it is already quite functional
<ssvb>
oliv3r: non-existing security is what made the original allwinner design so great in the first place :)
<oliv3r>
ssvb: i'm supprised; happily so
<oliv3r>
ssvb: i thought by rewriting the kernel side (v4l mem2mem basically) the whole vdpau lib would have to be rewritten
<oliv3r>
but if it uses the vdpau interface; won't it 'just plug into' xbmc?
<oliv3r>
wasn't that the definition of vdpau as interface?
<ssvb>
oliv3r: the current api of the cedar kernel driver is a major security hole by design, so the userspace code will have to be rewritten too
<oliv3r>
that part I knew
<oliv3r>
but because of that; I find the vpdau stuff we have, while useable, still PoC code imo
<ssvb>
oliv3r: libvdpau-sunxi implements the userspace vdpau api and is using the allwinner kernel module api as a backend, but I guess jemk can explain better :)
<oliv3r>
anyway, as for fosdem; in the end, it is 'just a PC' so besides demoing a desktop, vdpau i guess it's not too exciting
<oliv3r>
i should install my demo on my SSD, should make things really fast i'd think. experimentation time tomorrow!
<oliv3r>
my current SD card is class (4)
<nove>
oliv3r, ssvb, the kernel driver also needs to have some kind of scheduler to support multiples clients using it
<ssvb>
nove: that's a good point
<oliv3r>
multiple clients means multiple decoded video's
<oliv3r>
?
<oliv3r>
what about simultanious encoding and decoding, should be possible too?
<jemk>
oliv3r: i also thought about v4l mem2mem once, but i couldn't find any working userspace software using it
<nove>
oliv3r: with a libjpeg-sunxi made in a hurry, you could also show web browsing with hardware accelerated jpeg decoding
<oliv3r>
jemk: the mainline kernel has drivers using it, i think the exynos stuff is using it
<oliv3r>
jemk: i asked on the linux-media mailing list, what the best way was to start writing drivers for this, and they said 'easy, we have mem2mem for this purpouse'
<oliv3r>
dunno how that'll be performance wise
<jemk>
oliv3r: yeah, i know, but what player uses it???
<nove>
oliver, frame by frame it should be possible
<oliv3r>
nove: will it actually be visble? do we even know if libjpeg-sunxi would be faster then neon?
<oliv3r>
nove: in any case, yeah, libjpeg-turbo-sunxi ;p
<oliv3r>
nove: it's still nice to offload it from the CPU never the less
<oliv3r>
nove: even more interesting is jpeg encoder, thumbnailing depends on it imo
<oliv3r>
jemk: Well how does samsung do video decoding generally? do they have linux userspace or only android?
<oliv3r>
jemk: if only android, they just plug in the android lib -> mem2mem
<nove>
oliv3r: there is still no benchmarks, but as sample, this A13 encodes 4080x4080 faster, that my desktop can decode it
<oliv3r>
jemk: but for now, i wouldn't worry about it, once we start replacing cedarx.ko with cederus.ko, then we will see
<oliv3r>
nove: very sweet; well also, libjpeg would be used by mjpeg encoder right?
<oliv3r>
so there's def. uses for it
Night-Shade has joined #linux-sunxi
<nove>
nove: the jpeg type here is the only for mjpeg, only baseline i think, yes encoding and decoding
<jemk>
oliv3r: btw, its cedrus, latin for cedar (the tree), not cederus ;)
<oliv3r>
jemk: i noticed; mnemoc kept calling it cederus
<oliv3r>
jemk: but yes, cedrus!
<ssvb>
jemk: maybe we should also have look at gstreamer and hook the hardware video encoding/decoding there
<ssvb>
jemk: afaik mozilla is using gstreamer for the accelerated html5 video playback, we need to test if it works
Black_Horseman has joined #linux-sunxi
<oliv3r>
jemk: so how does userspace interact with vpdau? isn't it jsut a 'plugin' kinda thing? some glue code and go?
<jemk>
ssvb: i hoped to only write one standard api (like v4l mem2mem should be), i'm not really a software gui who wants to figure out a thousand different apis
<ssvb>
jemk: gstreamer should already have vdpau support
<ssvb>
jemk: I don't mean that it is you who has to do this work, it was just a general comment
<nove>
ssvb: jemk, isn't ffmpeg a better target?
<oliv3r>
jemk: well isn't it mplayer/xbmc -> libvdpau -> vdpauapi -> libcedrus -> cedrus.ko -> mem2mem -> silicon?
<oliv3r>
i thought libcedrus would talk ot the kernel via mem2mem
<ssvb>
jemk: does a standard kernel api exist? and does it allow to use the cedar hardware efficiently?
<oliv3r>
well if we have that, what's the big thing to make it ffmpeg -> libvdpau, and xbmc -> libvdpau
rz2k has joined #linux-sunxi
<oliv3r>
isn't it basically all the same?
<jemk>
ssvb: v4l mem2mem should do it, and is said to be efficient (zero copy possible)
<ssvb>
oliv3r: in the kernel we need some interaction between the cedar hw decoder and sunxi disp to pass pixel buffers without copying them
<jemk>
but i haven't looked too deep into it, as i can't find any player being able to use/test it
<oliv3r>
Turl: remember my nop question; it can't be by any chance be pipelining can it?
<jemk>
ssvb: v4l should do dmabuf
<oliv3r>
ssvb: i'm talking far future here though; when we have 'mainline quality' disp driver, and we want to do mailine quality cedrus driver, won't mem2mem suffice here too? (taking that arch into account?)
<oliv3r>
i don't think we should try to put too big an effort in fixing cedarX.ko for 3.4
<oliv3r>
what we have now, for 3.4 is 'fine'
<jemk>
for 3.4 i will definitely stay at the current state with vdpau, as i only see this as test environment.
<ssvb>
jemk: yes, we need to check what modern frameworks exist and whether they may be imposing any serious limitations
<ssvb>
oliv3r: for 3.4 the current insecure driver is just fine
<oliv3r>
yeah
<oliv3r>
ssvb: you are way more a graphics guy then I am; but cedrus would have to work even without X, right?
<oliv3r>
so the connection to disp, would have to be solved mostly in the disp (for mainline) at that time
<oliv3r>
i predict, that for mainline, disp will probably be the last driver ot get done
rellla has quit [Remote host closed the connection]
<wingrime>
oliv3r: we should use drm interface for mem2mem
<ssvb>
oliv3r: vdpau is a 'presentation' api, which is quite simple and nice for zero-copy playback
<oliv3r>
i really know to little about all this to be fair though :)
<ssvb>
oliv3r: zero-copy is implemented by simply feeding the physical buffer addresses to disp, and the job is done
<jemk>
and in mainline, as far as i understood it, one can use dmabuf for that, passing dma buffers between different drivers
<wingrime>
ssvb: sun9i have FE0-3 BE0-3
<wingrime>
err
<wingrime>
ssvb: sun9i have FE0-2 BE0-2
<ssvb>
oliv3r: if we have a universal encoding/decoding api, then we may have to struggle to get rid of redundant buffer copies
<wingrime>
ssvb: I mean FE0,FE1,FE2
<ssvb>
oliv3r: more complex and more generic api means that it might be more difficult to implement efficiently, or implement at all
<ssvb>
oliv3r: so I think that vdpau was a good choice for the start
<oliv3r>
we should be carefull what we call sun9i for now; as they changed the names a bit
<oliv3r>
well vaapi probably isn't too bad either :p
<oliv3r>
wingrime: do you have an allwinner device setup for mplayer decoding right now?
<mripard>
oliv3r: sunxi-babelfish is live :)
kivutar has quit [Remote host closed the connection]
<oliv3r>
mripard: after fosdem; i'll work more on sunxi-3.10, which should make babelfish really interesting and important
<oliv3r>
btw, I can now run power consumption tests on cb1 and cb2
<oliv3r>
(my cb3 has a battery, so won't work accuratly)
<mripard>
oliv3r: I'd like to support only the stable bindings
<wingrime>
oliv3r: sun9i have 1.5 MB SRAM (A1+A2+B+C)
<ssvb>
oliv3r: is it that tiny rtc clock battery?
<mripard>
(ie, the one that have been accepted to mainline.)
<ssvb>
oliv3r: or you got a real battery?
<oliv3r>
mripard: well 3.10 will get all the sunxi backported onto it; so that should be fine
<mripard>
otherwise, it will just be a maintenance hell
<mripard>
oliv3r: yeah, plus all the dirty bits
<mripard>
it's those that I'm not really fond of supporting.
<oliv3r>
mripard: the way i invision it; 3.10 will require both mechanisms anyway, 3.10+ drivers require a dtb to be loaded; < 3.10 drivers (e.g. 3.4 drivers) will require script.bin to be parsed from 0x430000)
<wingrime>
oliv3r: 200 kb A1+A2
<wingrime>
oliv3r: accoring memory map
FDCX_ has quit [Read error: Operation timed out]
<oliv3r>
mripard: i'm guessing, audio driver, disp driver, cedarx driver will be the main forwardported 3.4 drivers; rest should be mainline pretty easily
<oliv3r>
ssvb: an iPhone battery :) so yeah lipo
<oliv3r>
wingrime: so more sram
<wingrime>
oliv3r: thats according memmory map sizes
<wingrime>
oliv3r: not realy sure it fully correct
<wingrime>
oliv3r: BROM1 - 64 Kb
<wingrime>
oliv3r: BROM0 - 32KB
<oliv3r>
BROM0 BROM1?
<wingrime>
oliv3r: yes
<oliv3r>
i wonder what that is
<oliv3r>
i really want a device to poke at now :)
<wingrime>
sun6i looks like a80
<oliv3r>
for tablets, a23 is VERY interesting
<wingrime>
err
<wingrime>
sun9i looks like a80
Akaizen has joined #linux-sunxi
<wingrime>
wait a second, I calculate sizes for a23
jemk has quit [Quit: night]
<ssvb>
do we have any updates about a80?
<oliv3r>
10" IPS screen 1 GB A23 is a really nice (cheap) tablet
<oliv3r>
really depends on disp performance; i see all these tablets to 800x480, maybe due to the weakness of mbus/disp
<wingrime>
oliv3r: A1+A2 in sun8i -- 114kb
<wingrime>
oliv3r: 184320 A1+A2+B+C
<wingrime>
bytes
<bsdfox>
oliv3r, running that dd command did get my system booting from the sd card. I've got the kernel loading and have a rootfs on the card too but now I'm having a problem where my serial console on ttyS0 seems to stop working mid-way through boot http://pastebin.com/LW6Yhm5G
<oliv3r>
bsdfox: remember, you need a kernel with earlyprintk etc
<rz2k>
oliv3r: I mean A23 tablet link :D
<bsdfox>
oliv3r, 115200 seems to have done the trick. thanks
<bsdfox>
I would have expected garbage if it was just a baud rate issue
<oliv3r>
rz2k: oh that was my wish specs :p
<oliv3r>
rz2k: i'm a little weary about buying a 'quality' tablet, before i can see the screen; the TN panels i have are REALLY crap
<oliv3r>
bsdfox: depends, sometimes
<bsdfox>
I'm using linaro-server and find that wpa_supplicant is missing.. I compiled the kernel with support for my wireless adapter and it shows up with ifconfig -a.. is there a different rootfs I might look at that will allow me to get network access on this tablet?