<penguin42>
ManoftheSea: If you can get a board working with an android kernel, and you have a graphics driver then the real Linux userspace will be happy
<ManoftheSea>
neat.
bsdfox has quit [Ping timeout: 255 seconds]
<buZz>
hmmm, if i enable CONFIG_VIDEO_SUN4I_CEDAR, i get 'No rule to make target drivers/media/video/sun4i/cache-v7.c'
<buZz>
there seems to just be a .S there
<ManoftheSea>
that appears correct.
<ManoftheSea>
I have a cache-v7.S as well.
froese has joined #arm-netbook
<ManoftheSea>
as well as a sun4i_cedar.{c,h,o}
<buZz>
yeah .. why doesn't it compile then?
<buZz>
the makefile has cache-v7.o needed .. but nothing that says its a arm file?
popolon has quit [Quit: Quitte]
<ManoftheSea>
makefiles are a specific kind of magic... the kind I don't understand.
<ManoftheSea>
buZz: are you using "ARCH=arm CROSS_COMPILE=something"?
<buZz>
no, i am compiling native
<ManoftheSea>
I keep messing up my config by forgetting the ARCH one.
<ManoftheSea>
oh.
<buZz>
i'll just disable cedar for now ...
<froese>
ManoftheSea: then put it in the top-level makefile.
<ManoftheSea>
I'm not having a problem.
<nulluser>
ManoftheSea: Does that device really need the heat sink as pictured?
<nulluser>
I was going to buy one last night, ans scrolled down
<nulluser>
I think I spit up a little beer
<buZz>
i like the 400-pin odroid :P
gimli has quit [Ping timeout: 248 seconds]
<froese>
I don't understand why these things have to be that expensive
<nulluser>
The boards and component placement cost a lot in small quantities
<nulluser>
The BGA based devices need to be soldered in a reflow oven, or some other advanced technique
<servili007>
I'm pretty sure that programmers of past generations would be pretty ashamed of us considering these powerhouses expensive, given inflation, etc.
<servili007>
not to say that I'm going out and buying them all
<froese>
hmm... all smd boards go through a reflow oven
<ManoftheSea>
"that expensive" maybe because people buy them at this price?
<nulluser>
Not if you are hand assembling a batch of 50
<froese>
ok :-)
<froese>
but you won't hand-assemble a bga
<nulluser>
And I use a hot air station when I do smd work
<penguin42>
and getting BGA stuff right takes a bit of testing/fiddling
<nulluser>
I get get the small bgas to grap, like 32 ball
<nulluser>
*can
<nulluser>
yeah, it's voodoo
<froese>
not bad
<nulluser>
I don't know how it's even possible to get 100+ right
<ManoftheSea>
so here's where I'm stuck... I look at these, and I look at what I'm paying for a VPS right now...
<servili007>
nulluser: What station do you use?
<nulluser>
to be SURE they are all connected properly, and you can't really pass a current to test
<nulluser>
I have a Aoyue 968A
<nulluser>
it was very very cheap
<ManoftheSea>
Screw the virtualized host... Why can't I just get one of these ODroid-U2's in a colo for $10 a month.
<servili007>
nulluser: Yeah I've seen them around, do you like it?
<froese>
i guess a lot of testing with dummies
<penguin42>
ManoftheSea: So you need someone who can rack and stack a bunch of them together to take up 1 shelf in a datacentre
<nulluser>
It's okay, it's not a 2500 dollar station of course, but the regulation does work, and the air is the temp is says it is
<ManoftheSea>
penguin42: sorta like the EOMA-server, yeah.
<penguin42>
ManoftheSea: There are things like the baserock
<nulluser>
Little rough around the edges, loud, some glitches
<nulluser>
I bought the thermocouples and PID controller to make a reflow oven, but never got around to it.
<froese>
these calxeda boards/systems look nice but a different price range
<penguin42>
froese: Yeh but in terms of renting one card on them in a data centre, that probably makes more sense
<ManoftheSea>
yeah, like that, penguin42. it's just that these are a very clear performance per dollar (cpu, memory, disk space)...
<nulluser>
Also could not figure out how to keep the parts placed as a transfer it in. I was going to use a toaster oven, and you would have to have the hands of a surgeon to get it in without anything getting moved
<ManoftheSea>
Maybe a full virtualized x64 machine can do better... but I feel like I'm paying too much for a 64 MB VPS
<penguin42>
ManoftheSea: But to get them into a datacentre then you need to get the PSU sorted out for a rack, and remote admin of them
<nulluser>
I was looking at some OEM arm system on modules has night, some are pretty nuts
<ManoftheSea>
penguin42: I understand that there are costs I don't understand.
<nulluser>
I worked on a 3d engine on a pic. It had enough space for the framebuffer, but that's about it
<nulluser>
damn that's a nice display
<froese>
look at the video and the specs of the cpu *g*
<nulluser>
yeah that's really impressive
<penguin42>
yeh, although if you think the res of those displays is pretty much as the res we had on 8bit machines in the early 80s, and although the RAM is similar the clocks are faster
<nulluser>
I love that stuff, squeezing out every last drop of power from something
<nulluser>
Those guys were brilliant too ;)
<nulluser>
Look at something like super mario brothers 3 on a 6502
<froese>
2.5k RAM? the pet i started with had 8 *g*
<penguin42>
yeh I started on a BBC B with a massive 32k - although by the time you'd given 20k for screen RAM
<nulluser>
My first computer was a TRS-80.
<nulluser>
Came with a massive basic programming good, at about a 6th grade reading level
<froese>
the II was nice. but tandy was not were common in .de
<nulluser>
Kept me busy for years
<nulluser>
I loved those old machines because you _had_ to become clever to get them to do anything interesting
<nulluser>
You see these amazing assembly demos, and the best you can do in basic was to scroll so slow you could see the pixels move
<froese>
yeah. and you had to count every single byte.
<nulluser>
people are STILL writing C64 demos, to this day
<nulluser>
But we say that as we are hacking linux onto little android sticks ;) from the outside perspective we are all in the same category of nuts
<froese>
do they run it on a real C64? *g* i still have one in the attic but I think I will have problem to connect it to any monitor...
<nulluser>
composite is still there
<penguin42>
froese: Apparently, but they tend to do some stuff you couldn't do at the time, like precompute a load of stuff ona pc
<froese>
not here ... (only the video-in on the sat-card and that got borked in later kernels)
<nulluser>
shame.
stefanro has quit [Ping timeout: 260 seconds]
<nulluser>
they still sell composite stuff here, like the atari joysticks with an emulator and all the games
<penguin42>
and I think some cctv systems still use composite
<froese>
penguin42: right. even the programming is much more pleasent on a real computer. heck, assembling an 8k rom took an hour and you had to swap disks :-/
<penguin42>
froese: Yeh, and it's so much easier these days without having to put the UV on to wipe your EPROM
<nulluser>
heh
<nulluser>
I never trusted those things
<nulluser>
I would nuke the chip for an hour just to be double sure
<froese>
penguin42: i had a battery backed up ram for that *g*
<froese>
nulluser: they were ok. just check if they are ff and you're fine.
<nulluser>
Do you guys know of an SD emulator
<nulluser>
What I have in mind is a device that you plug into the SD card, and USB to the pc
<nulluser>
you can then mount it like a disk
<nulluser>
and read and write to it
<nulluser>
The actual disk image is a buffer on the PC, backed up to disk
<froese>
? plug into the card?!?
<nulluser>
plug into the female slot of the device you are developing for
robclark has quit [Excess Flood]
<froese>
hmm... nope.
<nulluser>
You could have the cross compiler output to it, and just reset the device. no messing around. Could also see log files in real time, etc
<froese>
well, that works with any sdcard *g* (i do it like that)
<penguin42>
nulluser: I've seen proposals for an SD card switcher that would sit in a slot and disconnect it from the host, but not for an emulator
<froese>
sorry, i compile on the testing machine.
<nulluser>
have to be a little more advanced than just a mass storage controller from the PC's perspective, otherwise you would need 4 gigs of ram on it
servili007_ has joined #arm-netbook
<penguin42>
nulluser: Well, you'd need 4GB of vm
<froese>
nulluser: if eth is working use tftp-boot and nfs-root.
<nulluser>
true, I would rather just have all of it on the host,except from some cache
<nulluser>
yeah, that's a little beyond me
<froese>
not difficult. easier than setting up u-boot *g*
<nulluser>
the devices I have a wireless only, so you would need a wpa_supplicant
servili007 has quit [Ping timeout: 245 seconds]
<nulluser>
that would get rought
<penguin42>
froese: There are homes for people who set up u-boot
<nulluser>
With just a natively supported Ethernet module I can see it
<froese>
nulluser: nfs-root over wlan would be hard, yes.
<penguin42>
well, no - not as long as you put the wlan setup in an initrd; most NFS roots these days are actually flips in initrd not actually direct nfs roots
<nulluser>
can uboot load over serial, with zmodem or some such
<penguin42>
yeuch
<nulluser>
That would be an improvement to what I have now
<nulluser>
Is that a sound of disdain, or a confirmation ;)
<penguin42>
I suspect it can - it sounds horrific though
<penguin42>
can you get jtag?
<froese>
penguin42: right. but try to get a wlan setup in limited environment of an initramfs.
<nulluser>
I don't know if it's exposed on the board
<froese>
if your u-boot has z-modem support, you can download the kernel via serial. but it will be slow...
<froese>
anyway. does anybody know, if there's some kernel work going on for an abstraction layer that allows drivers to be build for either PCI or OF interface?
<froese>
(in my case ethernet driver)
<penguin42>
hmm what hardware have you got that can appear as either - is it just that the firmware is setting the pci bridge up and only handing it out via OF?
<froese>
penguin42: there are chips with either PCI interface or directly embedded into a SoC (same mmio/irq etc, only config is different). i.e. via-rhine.
<penguin42>
oh ok, never come across that
<penguin42>
froese: Something like IDE perhaps? PCI-IDE is basically the same type of thing where the PCI part of it is just saying where the interface is
<froese>
one example are the via wondermedia chips (wm8x50).
<froese>
penguin42: exactly.
<froese>
or 16550 cards.
<penguin42>
froese: So I think you'd want to end up with 2 or 3 things, a core device, a pci- wrapped thing and a of wrapped thing; although I'm guessing OF can specify where PCI stuff is
<froese>
basically, the drivers only use pci function for getting register and irq settings and then never touch any pci function any more.
<penguin42>
froese: yeh, although watch out for power management stuff
<froese>
secondary problem *g*
<penguin42>
hehe yes
<froese>
in via's android kernel, they've gone the cheap way: they emulate a dummy pci controller and use the regular pci driver. but that pulls the complete pci stuff into the kernel :-/
<penguin42>
froese: I guess when the device has been around as a PCI device for years that's not an insane way to start
<froese>
as a start it may be ok. but they use it for years *g*
<froese>
and considering via's history i don't think they will change it.
<froese>
ever
<penguin42>
what's their insentive?
<froese>
hmm... better mainly kernel integration? but i guess that brings no money so they don't care.
<froese>
s/mainly/mainline/
<ibot>
froese meant: hmm... better mainline kernel integration? but i guess that brings no money so they don't care.
<penguin42>
yeh, although heck they could probably get a pci emulation like that in if it was strictly in a per-device section
<froese>
? what's that ibot?
<froese>
a mind reader? *g*
<froese>
penguin42: hmm... i don't know. it is a kludge and the kernel maintainers don't like that.
<penguin42>
yeh
dyoung-away is now known as dyoung
<froese>
maybe as a part of the devicetree integration. a dummypci with the devices to be connected configured via the dts.
<penguin42>
froese: It might be worth looking if there are similar things for AMBA, I seem to remember someone was asking what to do about AMBA devices connected to PCI but I guess those are actually bridged
XenGi is now known as XenGi_
<froese>
looking...
<penguin42>
froese: Hmm and maybe virtual devices?
<froese>
penguin42: ? what do you have in mind?
XenGi_ is now known as XenGi
<penguin42>
froese: I'm fairly sure in kvm that the disc, ether and video devices appear as pci, but I don't know if that's something hacked at the kernel level or at the emulator level
<froese>
hmm... not sure either. i'm only using the virtio stuff *g*
<penguin42>
froese: Well that's what I mean, the virtio stuff appears as pci
<froese>
right you are.
<penguin42>
probably a hack at the QEMU/bochs level
<froese>
no idea. never had a look.
sv has quit [Read error: Connection reset by peer]
sv has joined #arm-netbook
sv has joined #arm-netbook
sv has quit [Changing host]
freakazoid0223 has quit [Quit: Leaving]
bsdfox has joined #arm-netbook
beatnick_ has joined #arm-netbook
<froese>
penguin42: i looked at the amba/pci stuff. it's the other way around, an amba bus behind a pci bus.
<penguin42>
froese: Yeh, that's what I suspected
merbzt has quit [Read error: Operation timed out]
froese has quit [Quit: Leaving]
froese has joined #arm-netbook
penguin42 has quit [Quit: Leaving.]
beatnick_ has left #arm-netbook [#arm-netbook]
bsdfox has quit [Ping timeout: 255 seconds]
dyoung is now known as dyoung-away
aholler has joined #arm-netbook
froese has quit [Quit: Leaving]
aholler_ has quit [Ping timeout: 244 seconds]
XenGi is now known as XenGi_
AC`97 has joined #arm-netbook
<AC`97>
WarheadsSE: mooo
Quarx has joined #arm-netbook
hst__ has quit [Ping timeout: 245 seconds]
harrow has quit [Ping timeout: 246 seconds]
harrow has joined #arm-netbook
soooga has joined #arm-netbook
eFfeM has joined #arm-netbook
ManoftheSea has quit [Ping timeout: 246 seconds]
soooga has quit [Max SendQ exceeded]
soooga has joined #arm-netbook
ManoftheSea has joined #arm-netbook
soooga has quit [Max SendQ exceeded]
arete74 has quit [Ping timeout: 246 seconds]
arete74 has joined #arm-netbook
soooga has joined #arm-netbook
bsdfox has joined #arm-netbook
Skoti has joined #arm-netbook
Skoti has quit [Client Quit]
servili007_ has quit [Read error: Connection reset by peer]
Ershov has joined #arm-netbook
Ershov has left #arm-netbook [#arm-netbook]
rellla has joined #arm-netbook
RITRedbeard has quit [Quit: Leaving]
hipboi has joined #arm-netbook
ganbold__ has quit [Ping timeout: 245 seconds]
ganbold_ has joined #arm-netbook
gimli has joined #arm-netbook
torpor has joined #arm-netbook
<torpor>
morning
orly_owl has quit [Quit: leaving]
orly_owl has joined #arm-netbook
popolon has joined #arm-netbook
lkcl_ has quit [Ping timeout: 252 seconds]
HSt_ has joined #arm-netbook
rellla1 has joined #arm-netbook
orly_owl has quit [Quit: leaving]
rellla1 has quit [Client Quit]
rellla has quit [Ping timeout: 245 seconds]
rellla1 has joined #arm-netbook
rellla1 has quit [Client Quit]
Maqs has quit [Ping timeout: 272 seconds]
Maqs has joined #arm-netbook
bsdfox has quit [Ping timeout: 260 seconds]
orly_owl has joined #arm-netbook
rellla has joined #arm-netbook
rellla has left #arm-netbook [#arm-netbook]
rellla has joined #arm-netbook
<techn_>
mnemoc: Now I'm going for dynamic resolution part.. this is a bit trickier
<techn_>
but I have found couple bugs during this
rellla has quit [Read error: Connection reset by peer]
<mnemoc>
bugs where?
rellla has joined #arm-netbook
CaCtus491 has quit [Quit: leaving]
lkcl has joined #arm-netbook
cubiefan has joined #arm-netbook
<cubiefan>
hello
<cubiefan>
Is there anyone who have any 3d modeling file of cubieboard?
<torpor>
there is one on thingiverse
<torpor>
just search for cubieboard
<cubiefan>
oh thanks!
hg_5 has joined #arm-netbook
rellla1 has joined #arm-netbook
eFfeM has quit [Quit: Leaving.]
rellla has quit [Ping timeout: 251 seconds]
penguin42 has joined #arm-netbook
<techn_>
mnemoc: in stage branch.. dunno if they effect something.. but atleast inconstencies
tinti has quit [Read error: Connection reset by peer]
<ssvb>
rz2k: has page flipping ever worked correctly with xf86-video-mali in any configuration on A10 hardware?
<rz2k>
no idea, but MaliDRI2ScheduleSwap is called correctly
<rz2k>
I believe techn_ checked that
<rz2k>
he is also on our 'disp' team
<mnemoc>
rz2k: are you familiar with the edid stuff?
<rz2k>
nope
<mnemoc>
ok :)
<rz2k>
I have vga monitor connected to my mele, so never needed that
<mnemoc>
I bought an hdmi/vga converter and got some decent-looking edid debugging output <http://dpaste.com/855582/> and was wondering if it implies that the converter is good and that eventually I'll be able to use it ;-)
<mnemoc>
(the xrandr above is obviusly from my laptop)
<mnemoc>
it's not easy to find inexpensive real hdmi/vga converters
humoluden has joined #arm-netbook
<techn_>
mnemoc:
<mnemoc>
techn_: hey! please read the 3 lines above :)
<buZz>
mnemoc: its not?
<buZz>
i saw a lot of HDMIVGA convertors for ~40 euros
<buZz>
really lots and lots
<mnemoc>
that doesn't count as inexpensive
<techn_>
rz2k: thats new
<buZz>
well decoding HDMI and reencoding to VGA takes power
<buZz>
mnemoc: what did you pay for yours?
<mnemoc>
9
<buZz>
wow
<mnemoc>
so if it works... :)
<buZz>
it will be sold out quickly ;)
<mnemoc>
:)
<buZz>
i should be trying those kernel mods
<buZz>
to get proper res on my lapdock :)
<mnemoc>
buZz: no need to apply patches, it's all in the stage branches on the normal git repo
<buZz>
well, need to figure out which branch then :P
<mnemoc>
stage/sunxi-3.0 or stage/sunxi-3.4
<mnemoc>
depending on your taste
<buZz>
is 3.4 proper?\
<mnemoc>
it has some glitches... but some people uses it
<mnemoc>
3.0 is safer
<buZz>
ok, i'll stick to 3.0 for now
<mnemoc>
sunxi-wise they are most equivalent. except for some usb gadget and 8250 issues
<techn_>
mnemoc: what your laptops sysfs says?
<techn_>
mnemoc: 8250 issues.. do you know why they want to use custom driver for that?
<mnemoc>
techn_: they inject some busywaiting hacks
<techn_>
16520A worked well what I tested
<mnemoc>
techn_: my laptop doesn'tt have any decent graphics/fb*/modes. U:640x480p-73
<mnemoc>
(subject is mine, original commit has "more fixes" as message)
<mnemoc>
the idea of cherry-picking that makes me sick
<techn_>
+1
<ssvb>
rz2k: yes, MaliDRI2ScheduleSwap gets caled, but it has branches to handle 3 different cases, and branch which is supposed to do flipping is never used for me
<techn_>
is it always hw 32bit and converted by software?
<ssvb>
what do you mean, mali seems to be rendering in rgb format
<ssvb>
yes, it seems to be rendering to 32bit format, but software conversion to 16bit format is faster that 32bit->32-bit memcpy because it is memory bandwidth limited :)
<ssvb>
but there should be no copy or conversion operations if flipping is properly done
<techn_>
framebuffer has possibility to render 1 to 32bpp, but needs some work
<techn_>
you'll need to use that other layer instead of scaler layer.. I send that patch before
<mnemoc>
ssvb: adding g2d support will make x11 support MUCH better
<techn_>
mnemoc: but that wont work with a13 and a10s?
<mnemoc>
it's not clear if they have it or not
<techn_>
yep
<mnemoc>
but sure, it needs to be "smart" :<
<techn_>
DE_BE_HWC_CRD_CTL0x08d8? Bhardware cursor coordinate control register
<techn_>
DE_BE_HWC_FRMBUF0x08e016? Bhardware cursor framebuffer control
<ssvb>
mnemoc: yes, I'm primarily interested in g2d :) and just got sidetracked a bit because xf86-video-mali is simply not fit for the job
<mnemoc>
having basic code working on A10 we can confirm other sunxi SoCs have 2d accelerator or not, and in the worst case we will have awesome X11 for A10-only :)
<mnemoc>
ssvb: :)
<ManoftheSea>
mnemoc, you mean I should move from sunxi-3.0 to "staging"?
<mnemoc>
ManoftheSea: stage/sunxi-3.0 has improved (but wanting feedback) hdmi support
* penguin42
is kind of surprised that no one has ended up using hot nitric and an electron microscope to figure some of this stuff out
<mnemoc>
penguin42: if you have access to those things, please do :)
<penguin42>
mnemoc: I don't, but I don't think they're that rare, especially in universities
hno has quit [Excess Flood]
hno has joined #arm-netbook
<ManoftheSea>
mnemoc, I'm pulling the latest from amery/sunxi-3.0. Is that what you mean, or is there a different branch I should get.
<mnemoc>
ManoftheSea: that's the branch, yes.
<penguin42>
mnemoc: I remember my last company I worked for (a startup) a few years ago ended up diagnosing a fault on a chip they had made by hiring use of an ion beam/microscope for an afternoon or two