Rondom has quit [Remote host closed the connection]
Rondom has joined #linux-sunxi
p_rossak has quit [Ping timeout: 240 seconds]
florianH has quit [Quit: Connection closed for inactivity]
xes_ has joined #linux-sunxi
xes has quit [Ping timeout: 240 seconds]
Guest78775 has quit [Read error: Connection reset by peer]
xes_ has quit [Ping timeout: 256 seconds]
p_rossak has joined #linux-sunxi
xes_ has joined #linux-sunxi
p_rossak has quit [Ping timeout: 260 seconds]
p_rossak has joined #linux-sunxi
wzyy2 has joined #linux-sunxi
lkcl has quit [Ping timeout: 256 seconds]
apritzel has quit [Ping timeout: 240 seconds]
bingyun has joined #linux-sunxi
bingyun1 has joined #linux-sunxi
bingyun1 has left #linux-sunxi [#linux-sunxi]
bingyun has quit [Ping timeout: 260 seconds]
ninolein has joined #linux-sunxi
ninolein_ has quit [Ping timeout: 260 seconds]
bingyun1 has joined #linux-sunxi
egbert has quit [Disconnected by services]
egbert has joined #linux-sunxi
Andy-D has quit [Ping timeout: 264 seconds]
dave0x6d has joined #linux-sunxi
cnxsoft has joined #linux-sunxi
chatter29 has joined #linux-sunxi
<chatter29>
hey guys
<chatter29>
allah is doing
<chatter29>
sun is not doing allah is doing
<chatter29>
to accept Islam say that i bear witness that there is no deity worthy of worship except Allah and Muhammad peace be upon him is his slave and messenger
<chatter29>
to accept Islam say that i bear witness that there is no deity worthy of worship except Allah and Muhammad peace be upon him is his slave and messenger
chatter29 has quit [Client Quit]
reev has joined #linux-sunxi
velly has joined #linux-sunxi
pg12 has quit [Ping timeout: 260 seconds]
terra854 has joined #linux-sunxi
pg12 has joined #linux-sunxi
Hao_ has joined #linux-sunxi
victhor has quit [Ping timeout: 268 seconds]
laj has quit [Quit: Page closed]
book`_ has quit [Quit: Leaving]
lurchi__ has quit [Read error: Connection reset by peer]
lurchi__ has joined #linux-sunxi
book` has joined #linux-sunxi
TheSeven has quit [Ping timeout: 260 seconds]
TheSeven has joined #linux-sunxi
_whitelogger has joined #linux-sunxi
jernej has joined #linux-sunxi
DullTube has joined #linux-sunxi
iamfrankenstein has joined #linux-sunxi
bingyun1 has quit [Ping timeout: 256 seconds]
bingyun has joined #linux-sunxi
f0xx has joined #linux-sunxi
jernej has quit [Ping timeout: 240 seconds]
JohnDoe_71Rus has joined #linux-sunxi
reinforce has joined #linux-sunxi
alexxy has quit [Remote host closed the connection]
Hao_ has quit [Ping timeout: 268 seconds]
velly has quit [Read error: Connection reset by peer]
reev has quit [Read error: Connection reset by peer]
solarnetone has quit [Ping timeout: 240 seconds]
paulk-elm has joined #linux-sunxi
apritzel has joined #linux-sunxi
solarnetone has joined #linux-sunxi
apritzel has quit [Ping timeout: 258 seconds]
<silviop>
i can convert if allah give me tc1680j source code (or mali).
<KotCzarny>
that's cheap
<montjoie>
oh my god, Im doomed, dmiller just add a patch that destroy my revert patch for stmmac
solarnetone has quit [Ping timeout: 240 seconds]
alexxy has joined #linux-sunxi
<wens>
rebase and try again?
<montjoie>
too late
<montjoie>
the patch added patch the reverted patch:)
<montjoie>
so now I need to revert two patch
<montjoie>
didnt understand why dmiller do this
<montjoie>
since he agree to the revert
<montjoie>
and the breaker is totally silent since...
<KotCzarny>
conspiracy?
<montjoie>
yes probably
<plaes>
silviop: do you have any pictures of the tc1680j chip?
<montjoie>
at least, the good news is that khillman agree that I do the network test for kernelci
<KotCzarny>
you know what support xunlong provides?
lurchi__ has joined #linux-sunxi
<MoeIcenowy>
without a well community orange pi can only be nothing
<hojnikb>
if this can actually handle 2g connection and run a few scripts, it's good enough for me
<KotCzarny>
check if they have images/bsp for it
<KotCzarny>
if not.. happy hacking
paulk-elm has quit [Remote host closed the connection]
paulk-elm has joined #linux-sunxi
<KotCzarny>
but yeah, with working and stable software this might be a very good deal
<KotCzarny>
'it can run android 4.4, rrright, with 256M of ram ;)
<hojnikb>
if headless
<hojnikb>
is don't see why not
<plaes>
headless Android?
<KotCzarny>
turtloid
<hojnikb>
sure
<hojnikb>
this board isn't meant for displaying things
<petard_>
Camdroid runs on 64MB
<MoeIcenowy>
Camdroid is already not Android
<MoeIcenowy>
it's only an Embedded Linux based on bionic
<petard_>
Well technically it is :)
<KotCzarny>
i wonder if it comes with that 500mb nand populated with some working os
<hojnikb>
probably android
T0mW has joined #linux-sunxi
paulk-elm has quit [Remote host closed the connection]
paulk-elm has joined #linux-sunxi
victhor has joined #linux-sunxi
<KotCzarny>
googling revealed 0 (zero) possible linux softwares for that rda chip
<willmore>
2G? Well, that rules out my continent.
Hao_ has quit [Ping timeout: 240 seconds]
<KotCzarny>
might be my goo-fu failing.
<willmore>
Oh, Vivante graphics? So, there's some chance it'll get 3d support. ;)
chlorine has joined #linux-sunxi
<plaes>
good luck getting a datasheet...
<willmore>
30Mtri/s? 120Mp/s? For perspective, isn't that PlayStation 1 level graphics?
<MoeIcenowy>
plaes: but it's something cut down
<MoeIcenowy>
with only the first part
<MoeIcenowy>
and all hardware detail-related documents are removed
chlorine has quit [Remote host closed the connection]
<willmore>
This board looks like Xunlong had a customer that needed something specific and they later decided to sell it to others--or the customer dropped them and they had an unused design.
JohnDoe_71Rus has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org]
<MoeIcenowy>
maybe it's only because of some designs of RDA chips
<plaes>
yeah.. HW-wise it looks sweet.. but...
* willmore
doesn't get interested until 512MB
<KotCzarny>
willmore: buy two then ;)
<KotCzarny>
will double the cores too
<plaes>
:D
chlorine has joined #linux-sunxi
chlorine has quit [Read error: Connection reset by peer]
chlorine_ has joined #linux-sunxi
chlorine_ has quit [Remote host closed the connection]
<KotCzarny>
'orange pi 2g-iot is open source', lol? [citation needed]
cnxsoft has quit [Quit: cnxsoft]
serga has joined #linux-sunxi
<serga>
Hello! I'm looking for a SBC in order to make an IoT-oriented system (headless, autonomous and self-reliant).
<TheLinuxBug>
okay
<TheLinuxBug>
there are a lot
Ntemis has quit [Ping timeout: 268 seconds]
<serga>
So I think it's important to guarantee clean shutdowns and a battery could be the cheapest option (I'm a student with little budget).
<KotCzarny>
cannibalize some old allwinner a10/a20 based tablet
<KotCzarny>
cheapest and most things are already there
<TheLinuxBug>
^^
Nacho_ has joined #linux-sunxi
<serga>
I have found the OLinuXino, which seems to have a robust and reliable design, but are somewhat expensive compared to other brands.
<KotCzarny>
and people often just throw them out because of old android (and they can run mainline linux without a problem)
<serga>
Do you know other alternatives that have good software support?
<TheLinuxBug>
um what type of software support are you expecting?
<serga>
I'm a electronic noob, so maybe using an old tablet is not a realistic option for me, sorry.
<TheLinuxBug>
um
<serga>
Just running something like armbian.
<TheLinuxBug>
that would be all the more reason to learn wouldn't it?
<serga>
That's a good point.
<KotCzarny>
old tablet can be made into normal pc with ease, no need to learn electronics
<KotCzarny>
it just boots of sdcard
<KotCzarny>
and has usb, and wifi usually
<serga>
I'll get some information about it, thanks!
<willmore>
KotCzarny, LOL
<serga>
Maybe a C.H.I.P could be a good option? Someone have experience with that?
<KotCzarny>
you still didnt specify what will be the use case
<KotCzarny>
so compare old a10 tabled for 5usd with chip with all the addons for 20-30
<serga>
Datalogger of a weather station, sorry.
<serga>
Are there old A10 tablets for 5 USD?
<KotCzarny>
yup, second hand
<KotCzarny>
flea markets, etc
<KotCzarny>
sometimes even trash bins
chlorine has joined #linux-sunxi
<willmore>
serga, what kind of data are you logging?
chlorine has quit [Remote host closed the connection]
chlorine has joined #linux-sunxi
<willmore>
Data logging has three main parts (like most things): input, processing, storage.
<KotCzarny>
and fourth, sharing
<willmore>
For input, you need to make sure that you have the connections on the board (and the software to drive them) to speak to the devices from which you need to capture data.
<willmore>
Processing is generally data reduction/event capture.
<willmore>
Storage (sure, KotCzarny, maybe some sharing) is generally stuffing the data away somewhere that it can be further processed. This could be onboard storage, network storage, IFTT, etc.
<willmore>
If power consumption is an issue, than most SBCs will be poor choices as they use way too much power. You'll want to look into something like an Arduino or other board in those cases.
chlorine has quit [Ping timeout: 260 seconds]
chlorine has joined #linux-sunxi
<serga>
The main problem I encounter is to prevent light cuts, which are quite frequent in my area.
<KotCzarny>
are they longer than 3-5 hours?
<serga>
On the one hand because they make it difficult to collect data, and can corrupt the system (for weeks I will be 100 km from it).
<serga>
No, usually few minutes.
<KotCzarny>
then A series should do, as they usually have li-ion battery interface
<willmore>
Then a small battery should help out.
<willmore>
BenG83, were you working on the PB?
<KotCzarny>
cheap 18650 cell for 5-10usd will do, or again, reuse from old phone etc
<willmore>
Yeah, need a 5V boost/charger to work with that.
<BenG83>
willmore, kind of
<KotCzarny>
willmore, nope, axp does all that for you
<willmore>
BenG83, any chance that your work on the power management will benefit the Pine64?
<BenG83>
I have two prototypes atm
<willmore>
KotCzarny, sure, for the A series SoC.
<KotCzarny>
it will also charge the battery
<willmore>
I want to throw a Li-ion cell on my Pine-64 as a battery backup.
<BenG83>
I would like to help with a AXP803 mainline driver, I fixed some things on the BSP side, or rather figured out how battery parameters work
<serga>
So I should like for SBC with A series SOC (or any old tablet as suggested).
Hao_ has joined #linux-sunxi
<serga>
Thank you for your help.
<BenG83>
willmore, the Pine shop now has a 18650 cell holder
<willmore>
BenG83, I have plenty of cells. ;)
<BenG83>
for people outside US that can not order the LiPo
<BenG83>
I use that too atm for mine boards
<willmore>
I can also print my own cell holders. ;)
<BenG83>
are we talking about mainline or BSP?
<KotCzarny>
isnt axp autonomic?
<willmore>
BenG83, preferably mainline.
<willmore>
Not much love here for BSP--understandably.
<BenG83>
in any case, the AXP needs your battery parameters, and the registers need to be persistent
<BenG83>
in the BSP case you set battery parameters via dts
<BenG83>
and the Linux driver updates them only ONCE
<BenG83>
after that there are valid flags in the registers
<BenG83>
and the self learning process begins
<BenG83>
but if your capacitiy, RDC and 32 point LUT are off
<willmore>
How does one set the battery parameters in mainline?
<BenG83>
the battery gauge and charging will not work properly
<willmore>
Will they calibrate with a few charge/discharge cycles?
<BenG83>
I guess at the moment by writing the AXP registers directly
<BenG83>
in u-boot or ATF
serga has left #linux-sunxi [#linux-sunxi]
<BenG83>
since there is no driver yet
<willmore>
So, with some i2c peek/poke kind of thing?
<BenG83>
but I think MoeIcenowy made a small AXP daemone as a stopgap measure
<BenG83>
the AXP has two ways to handel battery parameters
<BenG83>
you can just write you capacity, RDC and 32 point LUT
<BenG83>
and it uses those
<willmore>
Is there a datasheet on it? I can go read that instead of bother you about this.
<BenG83>
or you can set the valid flags and then it tries to update the parameters over charging cycles
<BenG83>
sunxi wiki has the AXP803 datasheet
<BenG83>
or pine wiki
<willmore>
Okay, I'll get that and get a bit educated before I bother you further.
<BenG83>
I did some discharge tests on the PB
<BenG83>
the capacity is crucial to set right
<BenG83>
because if the fuel gauge is off
<BenG83>
most power mangagement daemons will shut down prematurely
<BenG83>
because they look at the % capacity reported
<BenG83>
not the voltage
<BenG83>
and voltage is tricky to interpret over short time frames
<willmore>
Makes sense.
<KotCzarny>
my banana pi r1 constantly tops off my 18650
<BenG83>
especially when your battery is large compared to the discharge current
wzyy2 has joined #linux-sunxi
<BenG83>
the discharge curve is flat
<BenG83>
and the voltage a bad indicator if you dont count charges
<willmore>
BenG83, is that info under the 'fuel guage' area?
<BenG83>
the AXP basically has self learning charge counters that get adapted from a start value
<BenG83>
yes
<BenG83>
so you enter battery parameters and then do a couple cycles
<willmore>
Yeah, I'll be using a large battery. 4Ah or better.
<BenG83>
the registers need to be persistent of course
<willmore>
That was the plan. :)
<willmore>
I'll look in the sheet how to do that.
<BenG83>
so you can´t disconnect the battery in between
<MoeIcenowy>
BenG83: the battery driver is still too far to go :-(
<willmore>
Or the registers get lost?
<BenG83>
yes
<BenG83>
or you have to keep DC IN
<willmore>
So, copy off the registers before poweroff and restore them on poweron?
<BenG83>
but if you connect all power you start from scratch
<BenG83>
that could work too
<willmore>
That's what android tends to do.
<BenG83>
there are some flags that indicate the sate of registers
<BenG83>
*state
<BenG83>
e.g. if learning is valid
<willmore>
Okay.
<willmore>
How does one talk to the chip without a driver? i2c?
<BenG83>
I would guess so
<willmore>
:)
<BenG83>
the ATF/u-boot can talk to it
<BenG83>
so you can just add your things there maybe
<BenG83>
that´s how I enable the charge LED output on the AXP
<BenG83>
there are some register read/writes in u-boot already to configure regulators
<willmore>
Gotcha. Having to think about uboot is beyond my knowledge.
<willmore>
but, since the chip does so much, I can see why uboot needs to control at least some of it.
hojnikb has quit [Quit: Page closed]
<BenG83>
the AXP are more or less specific to a SoC
<willmore>
Now it makes sense why Android distros keep a tiny filesystem with just the battery stats data on it--so that the little bootloader and the full OS kernel can share that data because one handles shutdown and the other handles startup and 'charge battery without booting OS' functions.
<BenG83>
yeah
<BenG83>
but if the registers are set ok on shutdown the AXP can do that on it´s own
<BenG83>
if you dont need a battery display on the screen :P
<willmore>
Assuming the battery doesn't completely die and the registers get lost...
<BenG83>
the AXP has two undervoltage lockout triggers
<BenG83>
one is for startup
<BenG83>
and one for online
<BenG83>
you can set them between 2.6 and 3.3V
<BenG83>
also different charge current limits for power budget management
<BenG83>
the PB charges with 1.5A when powered off
paulk-elm has quit [Ping timeout: 240 seconds]
<willmore>
uboot needs to look a the chip and see if it says the data is valid. If it does, then we're just restarting. If it's not valid, it needs to look on the little filesystem and restore that data--if it's valid.
<BenG83>
and with 0.5A when powered on
<BenG83>
the AXP also has a register to flag the cause of a poweron/poweroff
<BenG83>
so u-boot can check if it´s just a charger insertion or a powerbutton press
<willmore>
Interesting. That's based on what your expected power supply and SoC consumption are?
<willmore>
Okay, those flags would be handy.
<willmore>
But, honestly, those *or* the valid flags would be enough.
<BenG83>
the Pinebook is supplied with 5V/3A for example
<BenG83>
so you need to determine how much charge current you can spare while offline/online/suspend
<BenG83>
and reprogram the limits
<BenG83>
some tablets do that load dependend via a look up table
<willmore>
Too bad you can't say "input is 3A, give the battery what's leftover after powering other stuff".
<BenG83>
there are some PMIC that do that
<willmore>
The good ones. ;)
<BenG83>
I think AXP can handle 4A total
<willmore>
Thanks, BenG83.
<BenG83>
np
paulk-elm has joined #linux-sunxi
The_Loko has joined #linux-sunxi
engideavr has quit [Quit: Konversation terminated!]
<willmore>
That's an almost perfect looking chart for a constant wattage discharge.
<willmore>
Looks like you can shutdown at 10Ah and have a safe margin for the cells health.
<KotCzarny>
yeah, for constant wattage, i wonder what would happen if user would drop load on and off in between
<willmore>
KotCzarny, for a first order capacity calculation, that's not idea.
<KotCzarny>
ahm, ok
<KotCzarny>
so it's a calibration run?
<willmore>
Once you have some data about the current draw/voltage at various capacity points, then you can start being mean to the charge controller. ;)
<willmore>
KotCzarny, it was a 'can I get the controller to stop under reporting battery capacity' run.
<willmore>
Now, the controller should have a better idea as to the true capacity.
<willmore>
Different batteries will have different voltages at different current draws at different points along their discharge curve.
<willmore>
Ideally, you'd have a set of curves for different discharge rates.
<KotCzarny>
yeah, i know, higher internal resistance even with age
<willmore>
So, the charge controller could say "Okay, the load has X and I think the capacity is Y, so the voltage should be Z." So it can do sanity checks on all of those values.
<willmore>
KotCzarny, exactly, that, too.
<willmore>
But also with charge level.
<willmore>
Different chemestries have different curves for that, too.
<KotCzarny>
i've always wondered why battery controllers dont account for that in laptops (unless they did ~2005-2008)
<willmore>
I don't know how the charge controllers model that. They may try to calculate a set of R values for the internal resistance at varios charge levels.
<willmore>
KotCzarny, they may have, but that's maybe a more modern feature.
<KotCzarny>
imo they might have been just counting charge, and not bothering with curves
<willmore>
That's not a bad way to do it.
<KotCzarny>
bad if you need to report reliable % with aged batteries
<willmore>
The first chips were more meant just as a battery guage and Li-ion cells are pretty easy to gauge if you measure their Coulemetric charge(sp?).
<willmore>
Yes, you need to keep calibrating the full charge value. If you don't do that, you're going to be off. And for that, you need a full discharge every now and then.
<willmore>
Which users specifically try not to do because they don't like their devices shutting down on them with little warning.
<KotCzarny>
and my experience with thinkpads confirms that ('please run a full cycle to callibrate gauge')
<willmore>
If you'll remember, BIOSes back then often had battery maintenance menus where you could say "okay, charge the battery fully and then turn off charging and run yourself dry".
<KotCzarny>
but as i've said, my thinkpads are from 2005-2008
<willmore>
LOL
<willmore>
We are agreed.
<KotCzarny>
*calibrate
scream has joined #linux-sunxi
<KotCzarny>
do you know what is the current state of laptop/battery controllers? (apart from open and hackable ones from allwinner land)
<KotCzarny>
did they get smarter?
cptG_ has joined #linux-sunxi
LargePrime has quit [Ping timeout: 240 seconds]
Andy-D has joined #linux-sunxi
Nacho__ has joined #linux-sunxi
cptG has quit [Ping timeout: 260 seconds]
jernej has joined #linux-sunxi
Nacho_ has quit [Ping timeout: 258 seconds]
<diego71>
modern batteries have a seial protocol to check the status of the battery
<TheLinuxBug>
ya and some batteries actually come with their own circutry to manage charging them selves
<diego71>
KotCzarny: ^
<KotCzarny>
yes, but do they calculate their curves? communicating is one thing, but internals might be varying much
<TheLinuxBug>
In those cases I would assume the circutry included is catered to that specific battery and handles things and just communicates with the machine
<diego71>
depends on battery manufactor
<diego71>
10 years ago most of them use a coloumb counter
<diego71>
and mantain an estimate of total capacity than is updated overtime
<diego71>
i don't think now is much different
<KotCzarny>
so they just pretend to be smart then. BenG83: dont forget to file a patent, just in case ;)
<BenG83>
I think modern smart batteries like in laptops just have a multi-dimensional LUT
<BenG83>
that gets calibrated with a lot of measurements
<BenG83>
over temperature, cycle life, load, etc
<diego71>
charge level is a pain in the ass :) or so i was told
freemangordon has quit [Read error: No route to host]
<KotCzarny>
yup, everything flows
<KotCzarny>
cant charge the same battery in the same way twice ;)
<diego71>
for example voltage level depends on charge level but much more on temperautre and discharging current
freemangordon has joined #linux-sunxi
tkaiser has joined #linux-sunxi
LargePrime has joined #linux-sunxi
<diego71>
BenG83: it's possible but probably is also a trade secret
<KotCzarny>
diego71: most likely found in some patent application
<KotCzarny>
ie. details
<BenG83>
basically you would also have to protocol events like deep discharge that irreversibly change your internal resistance
<BenG83>
or overcurrent situations
<diego71>
probably with some details missing :)
<BenG83>
and then adapt again
<miasma>
tkaiser: the list here is already missing the opi zero h5 model :) https://forum.armbian.com/index.php?/topic/1351-h3-board-buyers-guide/
<jernej>
MoeIcenowy: I received monitor and also found the solution :)
<jernej>
MoeIcenowy: Issue is switching between DVI and HDMI mode
<jernej>
MoeIcenowy: I will first wait that main patches land and then send fix
<miasma>
are the arm64 instructions twice as wide too
reinforce has joined #linux-sunxi
matthias_bgg has quit [Quit: Leaving]
<MoeIcenowy>
jernej: oh I remembered the monitor is shown as DVI when using old patchset
<MoeIcenowy>
tkaiser: when will you make stable versions for H3/5 mainline image?
<willmore>
Why would an h5 be in the h3 buyers guide? ;)
lurchi__ has quit [Ping timeout: 264 seconds]
<wens>
v8 is written in C++, likely with extensive usage of virtual functions
<willmore>
Unless the h3 guide now just reads "The H5 is out, stop buying H3 boards." ;)
<wens>
add to that syntax parsing + JIT, means a lot of memory allocations + pointers
<tkaiser>
MoeIcenowy: No idea, I'm not involved in the 64-bit Armbian game (always just asking stupid questions :) )
<jernej>
MoeIcenowy: but I still don't know why do you dislike it. Seems pretty portable, IPS, capacitive touch... great for development :)
<willmore>
miasma, the instruction in 32 bit ARM can be 32 bit (native) or 16 bit (thumb2 mode). Instructions in 64 bit mode are all 32 bit.
<MoeIcenowy>
jernej: the only problem is that you need a graphics driver with full EDID capbility to use it
<MoeIcenowy>
tkaiser: who's involved? and how about H3? ;-)
<tkaiser>
wens: Yep, and the funny thing is that we maybe now start to understand user reports of arm64 Chromium/Firefox running unstable on Pine64, ODROID-C2 and others. Maybe those are simply 'swapping to death' already ;)
Hao has quit [Remote host closed the connection]
Hao has joined #linux-sunxi
lurchi__ has joined #linux-sunxi
naobsd has quit [Remote host closed the connection]
<wens>
very possible
<tkaiser>
MoeIcenowy: IgorPec and Mikhail (joinin IRC via backlog ;) )
<MoeIcenowy>
oh I seems to haven't see the name Mikhail many times...
<wens>
probably only reads the logs and never actually joined the channel :p
<MoeIcenowy>
tkaiser: in fact on my own distro nearly no browsers are usable under arm64
<BenG83>
I run that on your mainline branch with the Pinebook
<BenG83>
much better than firefox:armhf at least
<tkaiser>
BenG83: But Pinebook is a memory monster. 2GB are huuuuuge ;)
kloczek has quit [Remote host closed the connection]
naobsd has joined #linux-sunxi
<miasma>
tkaiser: i guess the text about opi zero plus 2 is also wrong here https://forum.armbian.com/index.php?/topic/1351-h3-board-buyers-guide -- 2xusb via gpio and one usb otg/power port, doesn't it make it 1+2, not 1+1+2
<jernej>
MoeIcenowy: I intent to improve U-Boot EDID situation. However, today I will try to dig up missing info about DE2 so I can give you more constructive replies to the driver.
<parazyd>
anyone with a lime2 here?
<MoeIcenowy>
jernej: ;-) thx
<parazyd>
i 'm trying serial, but it outputs gibberish, and the red led is pulsating
<parazyd>
yesterday it worked... i'm not sure what's up
<diego71>
parazyd: power problems?
<tkaiser>
parazyd: Undervoltage/undercurrent. Choose one or two ;)
<diego71>
how much pulsating
<parazyd>
oh you mean i should try more amps?
<parazyd>
diego71: pulsating constantly, in one second intervals maybe?
<diego71>
parazyd: check thr voltage
<tkaiser>
miasma: One OTG (that can be turned into a full host port), one USB receptacle and two on the headers?
<parazyd>
diego71: what do you mean by that?
kloczek has joined #linux-sunxi
<tkaiser>
parazyd: voltage drops are likely if something's wrong with the cable or your PSU is dying.
<miasma>
tkaiser: maybe i'm missing something, but i only see the headers and the otg port in the picture
<MoeIcenowy>
tkaiser: at least the OTG on OPi Zero have broken Vbus
<parazyd>
tkaiser: ah ok, i was thinking about serial voltage
<MoeIcenowy>
it cannot output power
<tkaiser>
miasma: Yep, you're correct. Will update it in a week
<MoeIcenowy>
only able to input
<MoeIcenowy>
despite it have id pin
<MoeIcenowy>
it's the reason why I set dr_mode to peripheral in my patchset which will land in 4.12
<KotCzarny>
parazyd: even with full power cycle and without any cables connected?
akaWolf has joined #linux-sunxi
<parazyd>
give me a minute to find a better psu
|Jeroen| has joined #linux-sunxi
tlwoerner has quit [Ping timeout: 240 seconds]
tlwoerner has joined #linux-sunxi
tlwoerner has joined #linux-sunxi
<MoeIcenowy>
tkaiser: you still didn't answer one of my question -- when will Armbian stables with mainline kernel be out for H3 boards?
lkcl has quit [Ping timeout: 260 seconds]
JohnDoe_71Rus has joined #linux-sunxi
<parazyd>
gibberish
<parazyd>
i have 0.5amp psu
<KotCzarny>
;)
<parazyd>
can try at home with 2amp
<KotCzarny>
some high powered usb port?
<parazyd>
the red led pulsates, 0.5seconds
<parazyd>
let me try that
<parazyd>
that i have two amps actually here
<parazyd>
(was using the barrel now)
<tkaiser>
MoeIcenowy: Don't know. :)
BenG83 has quit [Quit: Leaving]
petard_ has quit [Remote host closed the connection]
<parazyd>
yes
<parazyd>
KotCzarny: it works this way :)
<parazyd>
thanks!
Mr__Anderson has joined #linux-sunxi
<KotCzarny>
thank tkaiser for suggestion
chlorine has quit [Remote host closed the connection]
chlorine has joined #linux-sunxi
chlorine has quit [Read error: Connection reset by peer]
chlorine has joined #linux-sunxi
chlorine_ has joined #linux-sunxi
<miasma>
willmore: so it's interesting how they've paired 512MB of RAM with wider instruction sets in the cheap boards. maybe they're not such an issue with caches, but the amount of RAM is quite ridiculous
<miasma>
intel used to have old dual-core boards before core2 that actually performed worse with x86-64 code due to having only 1MB of cache per two cores
<KotCzarny>
i'm still amazed by the two groups of people, first: expecting lots of mem/performance out of sub-10usd boards, second: looking at the ram specs but actually using 1/10 of that
chlorine has quit [Ping timeout: 268 seconds]
<MoeIcenowy>
tkaiser: currently many people still do not dare to use the mainline images as they're claimed as "nightly"
muvlon has joined #linux-sunxi
<miasma>
KotCzarny: it's not about high expectations. the designs can be so bad that even mediocre perf is hard to achieve
<miasma>
the whole point of replacing h3 with h5 makes no sense if you still need to stick with old instructions
<KotCzarny>
well, neon unit is faster, for one
<KotCzarny>
so its worthwhile(?) just for that alone
<tkaiser>
MoeIcenowy: As I understood the others we still wait for stuff to stabilize on H3 as well before mainline images are released officially. Many H3 users are somewhat special. They spend $7 on a SBC and think they're entitled to get free support and superiour software in exchange.
<miasma>
tkaiser: indeed it really depends on the benchmark. some real world interactive applications might also benefit from a lower memory footprint
<miasma>
what kind of stabilization can be expected on h3? doesn't it work quite well already?
<KotCzarny>
is dvfs stable already?
<tkaiser>
miasma: For my use cases it runs well since decades. The average user will ask 'What's about video?! Mali drivers?!'
<miasma>
mainline ethernet/usb otg would be #1 on my list since it allows running other distros than armbian (without patching) if someone wants more out-of-the-box binary experience
<KotCzarny>
armbian-malivideo64-edition
chlorine_ has quit [Remote host closed the connection]
chlorine has joined #linux-sunxi
tkaiser has quit [Read error: Connection reset by peer]
chlorine_ has joined #linux-sunxi
Guest89155 has quit [Ping timeout: 240 seconds]
<muvlon>
in the old days, when you bought a device with a GPU, you could expect GPU support
<muvlon>
this didn't really change until the last few years
Keziolio has joined #linux-sunxi
Keziolio has joined #linux-sunxi
Keziolio has quit [Changing host]
<KotCzarny>
muvlon: not true
<KotCzarny>
see windoze/linux situation
<muvlon>
you could expect your hardware to work with *some* OS
<muvlon>
these days, you can buy a board with devices on it that even the OEM themselves can't figure out how to use
<KotCzarny>
have you ever bought 'development board pc' ?
chlorine has quit [Ping timeout: 260 seconds]
<muvlon>
yes
<KotCzarny>
set top boxes come with some software installed
<KotCzarny>
but sbc dev boards are just that, piece of hardware waiting for orders
<muvlon>
x86 devboards or console dev kits were never in the laughably bad state of support that some arm boards are
<hanetzer>
ja. arm board support packages are jokes at best.
<muvlon>
of course not everything works 100% with a dev kit, but it's not close
<muvlon>
some of these manufacturers ship their shit seemingly as soon as the first LED lights up
<KotCzarny>
i think the point here is actually lack of docs, not software
<hanetzer>
yeah. lack of docs is killer.
<muvlon>
that too
<hanetzer>
I've got some 29 or so kernel modules with very little in the way of docs as to how they actually work.
<KotCzarny>
and i tell you, it wont get any better, blackboxes. blackboxes everywhere. in your lap, in your car, in your fridge, in your dog.
|Jeroen| has quit [Quit: dada]
<hanetzer>
would that I could afford ida and maybe make some headway on this.
<KotCzarny>
and you know what? blackboxes in your blackboxes!
<jernej>
MoeIcenowy: Of course, but after I understand perfectly, which may take few days
lurchi__ has quit [Ping timeout: 260 seconds]
Andy-D has quit [Ping timeout: 240 seconds]
lurchi__ has joined #linux-sunxi
<tkaiser>
apritzel: Do you at ARM have some information/slides/thoughts on memory consumption on Aarch64 devices? I just ask since I noticed a huge difference between a NodeJS benchmark running with arm64 vs. armhf binaries today
<muvlon>
going from x86 to x64, you can expect an increase of 30% in binary size in some cases
<muvlon>
it depends on what portion of your code is pointers
<muvlon>
i'd bet it's about the same with armv8-A
<muvlon>
but in your case, it's almost 100%!
<miasma>
but, binary size on disk != binary size in ram
<muvlon>
they're not the same, but related
<muvlon>
size in ram is apparently even worse than size on disk
<miasma>
the relation is pretty loose if the application generates or loads lots of data with pointers
<tkaiser>
I don't care about storage but only RAM 'consumption' and strategies to deal with that.
<ssvb>
tkaiser: I would guess, it should be more or less the same as x86 vs. x86-64
<miasma>
e.g. some app that generates linked lists with a single integer payload would look really bad with fatter pointers
<miasma>
but isn't thumb code more compact than x86 asm
<miasma>
on average
<ssvb>
except if the minimal page size is larger than 4K on ARM64
<ssvb>
then ARM64 would be more wasteful
Ntemis has joined #linux-sunxi
<tkaiser>
Hmm... this needs a few considerations and most probably also time consuming testing. For example how zram might behave (faster code running in memory issues but compensating through zram).
<KotCzarny>
would it require app rewrite to test? or just recompilation in arm64 mode?
<KotCzarny>
tkaiser, recompile ioperf and check ;)
<tkaiser>
KotCzarny: First I need a way to tell Pine64+ to hide 512MB DRAM. No 64-bit board around with only 512MB DRAM.
<KotCzarny>
um, boot linux with mem=512 param?
lurchi__ has quit [Ping timeout: 264 seconds]
lurchi__ has joined #linux-sunxi
<KotCzarny>
another way would be hacking uboot memdetect code to return 512megs, but check that linux kernel standard mem param
<muvlon>
just run something that hogs 512MB and make sure it doesn't get oomkilled
<muvlon>
can you put a negative value into /proc/pid/oomadj? :P
<KotCzarny>
why, it's better to make linux limit itself to 512megs
<willmore>
miasma, 512KB on an AARCH64 processor is silly. 1GB is pretty tight, even. So many phones with 1GB and A53 cores. All running in 32 bit mode because they don't have enough memory to run 64 bit mode. *sigh*
<muvlon>
A53 on its own is a weird choice anyway
lurchi__ has quit [Ping timeout: 240 seconds]
<muvlon>
as a "little" core it makes sense but if you just put 8 A53 cores and nothing else what's the point?
<miasma>
at least I'm quite happy with the stuff they put in. the price/perf is right
<miasma>
yesterday there was one guy on another channel pondering which board he should buy. he probably ended up with a 100 eur rpi3 starter kit
<willmore>
muvlon, I agree that dual A53 clusters is pretty crazy. Better off with a cluster of A53 and dual A57.
florianH has quit [Quit: Connection closed for inactivity]
lurchi__ has joined #linux-sunxi
<willmore>
ssvb, the page size is the same in both modes, isn't it?
<miasma>
people pay 100 euros for rpi3 and complain how $9 opi quad core is too slow :P
<muvlon>
A57 was mostly ignored by chip makers for some reaso
<willmore>
Yeah, 4k.
<KotCzarny>
too power hungry for mobile soc?
<muvlon>
i'm just waiting for A72s and A73s to hit the low-end :)
<KotCzarny>
i'm happy with my banana (the original one) and oranges
<willmore>
The A57 was not a success. This came up a few weeks back over on #odroid when the new nVidia chip shipped as it has some of them in it.
<miasma>
would 10nm SoC process decrease the power consumption that much? would it be possible to combine dram and soc on the same chip?
<willmore>
Pretty much the only A57s that shipped were in three Samsung chips. That's pretty much it.
<muvlon>
the newest nvidia SoC is still using A57! and four of it even!
<KotCzarny>
maybe it was too expensive?
<muvlon>
and it uses them as "little" cores :P
<willmore>
The loss of 20nm at TSMC killed it as ARM partnered with TSMC to have A57 on 20nm at launch and 20nm never launched....
<ssvb>
willmore: the hardware should support 4K and 64K pages, but some kernels may be configured to only use 64K pages
<muvlon>
KotCzarny, yeah, I was thinking maybe they had already licensed that but didn't want to license the A72
<willmore>
So we were left on 28nm and A57 wasn't compelling.
<willmore>
ssvb, that would be crazy. It can do 4k, 16k, and 64k. I can't imagine using the larger sizes exclusively.
<willmore>
Okay, for data center/HPC workloads, it could make sense.
<muvlon>
thunderx is very far from anything that's in a phone, though
<willmore>
But, doesn't ARM have a large page ability like Intel?
<ssvb>
sticking to just a single page size makes the software implementation easier
<willmore>
Yeah, true, but the reason to use the large pages isn't software simplicity, it's because you're using large data sets and killing your TLB and TLB cache.
<muvlon>
I'm thinking maybe if you fab it in 14nm, the A57 actually becomes a quite power-efficient mid-range core
<muvlon>
and that's why nvidia are using it as their little core
<willmore>
muvlon, Actually the A72 followed on the A57 and is faster and more power efficient. It would be crazy now days to use the A57 for anything.
<ssvb>
using "one page size that fits them all" approach simplifies the software implementation (the kernel)
<muvlon>
willmore, can you fit an A72 in the same die space as an A57?
scream has quit [Remote host closed the connection]
<ssvb>
mixing and matching multiple page sizes may be complex and buggy
<willmore>
muvlon, yes. smaller and faster and more power efficnet.
<muvlon>
ah, then nvidia are probably just cheapskating
<willmore>
muvlon, the theory is that the chip design is actual several years old and that noone wanted it.
apritzel has quit [Ping timeout: 258 seconds]
<willmore>
ssvb, it's an engineering tradeoff. Do you want to lose the possible memory allocation efficiency of 4k pages vs the cost of implementing them?
<maik__>
Since some time, i am working on having analog audio out of my nano pi neo (REV 1.1) with using the on board LINE OUT pins. Should this work with just using nanopi_neo_defconfig and the sun8i-h3-nanopi-neo device tree? Last time I got the answer, that i have to patch one of the dts file. Didn't work. But why is this needed? The LINE OUT is availbale on the NEO and on the NEO2 as well. So this might be the default? (Some Logs: htt
<willmore>
If you need large data sets, then you have to support the 64k pages. Supporting 4k ones as well has to be judged on if they're worth it.
<willmore>
Hit enter once and a while.
foxx has quit [Ping timeout: 256 seconds]
<tkaiser>
apt install zram-config
<tkaiser>
Oops :(
<KotCzarny>
too much irc
<willmore>
That's not a bad idea, though.
<tkaiser>
willmore: Time to build a new kernel: # CONFIG_ZRAM is not set
<KotCzarny>
make it a module
<miasma>
you probably also want zrem & lz4, not lzo?
<miasma>
*zram
<willmore>
tkaiser, aww, bummer.
<tkaiser>
miasma: As I understood you can choose even at runtime? /sys/block/zram*/comp_algorithm
lurchi__ has quit [Ping timeout: 240 seconds]
lurchi__ has joined #linux-sunxi
iamfrankenstein has quit [Read error: Connection reset by peer]
iamfrankenstein1 has joined #linux-sunxi
Hao__ has quit [Ping timeout: 260 seconds]
vpeter has quit [Ping timeout: 258 seconds]
cosm has quit [Ping timeout: 258 seconds]
iamfrankenstein1 is now known as iamfrankenstein
vpeter has joined #linux-sunxi
cosm has joined #linux-sunxi
<miasma>
tkaiser: yes but you still need to compile in the support afaik
<KotCzarny>
so, back to our conversation from year ago that mem bw matters? ;)
<tkaiser>
It matters with certain workloads, sure. Especially since currently in Armbian we downclock DRAM on those small H3/H5 boards heavily by default.
<BenG83>
how far down?
<TheLinuxBug>
Is that for better stability or?
<tkaiser>
BenG83: 408 MHz by default. Consumption decrease for whatever reasons (don't understand it)
<KotCzarny>
TheLinuxBug: remember those tiny boards often come with microusb power