<FrostyBytes>
well several more binary searches later the complete list of modules needed for usb keyboard to work is: i2c-mv64xxx axp20x-i2c axp20x_usb_power phy-sun9i-usb sunxi {ehci,ohci}-platform hid-generic usbhid
<KotCzarny>
why are you doing it hard for you?
<KotCzarny>
also, sun9i is not for a20 afair
<KotCzarny>
a20 is sun7i
<KotCzarny>
though otoh i have it enabled too
<KotCzarny>
why dont you just use sunxi_defconfig?
ErwinH has quit [Ping timeout: 264 seconds]
libv has joined #linux-sunxi
<FrostyBytes>
I prefer to have a minimal kernel with everything that can be built as a module as a module
libv_ has quit [Ping timeout: 260 seconds]
libv_ has joined #linux-sunxi
libv has quit [Ping timeout: 250 seconds]
camh has joined #linux-sunxi
libv_ is now known as libv
foxx_ has joined #linux-sunxi
reinforce has joined #linux-sunxi
IgorPec has joined #linux-sunxi
msevwork has quit [Ping timeout: 246 seconds]
msevwork has joined #linux-sunxi
jernej has joined #linux-sunxi
alsy has quit [Ping timeout: 250 seconds]
jernej has quit [Ping timeout: 252 seconds]
ErwinH has joined #linux-sunxi
montjoie_ has joined #linux-sunxi
montjoie has quit [Ping timeout: 258 seconds]
ErwinH has quit [Ping timeout: 252 seconds]
ganbold_ has quit [Remote host closed the connection]
ganbold has joined #linux-sunxi
muvlon has quit [Ping timeout: 258 seconds]
muvlon has joined #linux-sunxi
FrostyBytes has quit [Ping timeout: 260 seconds]
dave0x6d has quit [Quit: Connection closed for inactivity]
xes has quit [Quit: WeeChat 1.6]
ErwinH has joined #linux-sunxi
cnxsoft has joined #linux-sunxi
akaizen has quit [Read error: Connection reset by peer]
akaizen has joined #linux-sunxi
jrg has quit [Ping timeout: 256 seconds]
<msevwork>
why do i have some word highlighted all the time in vim :) how do i disable that..tried F4 but didn't work
florianH has joined #linux-sunxi
perr has joined #linux-sunxi
jrg has joined #linux-sunxi
leviathanch has joined #linux-sunxi
montjoie_ is now known as montjoie
bugzc has quit [Ping timeout: 260 seconds]
my123 has quit [Ping timeout: 252 seconds]
my123 has joined #linux-sunxi
my123 has joined #linux-sunxi
my123 has quit [Changing host]
apritzel has joined #linux-sunxi
Worf has joined #linux-sunxi
leviathanch has quit [Read error: No route to host]
leviathanch has joined #linux-sunxi
LargePrime has quit [Ping timeout: 250 seconds]
popolon has joined #linux-sunxi
jrg has quit [Ping timeout: 258 seconds]
LargePrime has joined #linux-sunxi
jrg has joined #linux-sunxi
Putti has quit [Quit: Leaving]
_whitelogger has joined #linux-sunxi
xes has joined #linux-sunxi
fkluknav has joined #linux-sunxi
Putti has joined #linux-sunxi
Mr__Anderson has joined #linux-sunxi
jrg has quit [Ping timeout: 258 seconds]
<wens>
apritzel: does your u-boot h5 branch contain the latest a64 patches?
<wens>
the latest posted version i mean
<apritzel>
wens: yes
<apritzel>
I posted HEAD~something of that branch
<apritzel>
wens: do you want to port the R40 DRAM support on top of that?
<wens>
yup
dh1tw has joined #linux-sunxi
<ErwinH>
Setting up a testscript to see at what cpu-voltage the kernel crashes :)
jrg has joined #linux-sunxi
<KotCzarny>
dont forget cpuburn
<ErwinH>
Don't forget to sync the disk before the cpu crashes :)
<KotCzarny>
make a ramdisk?
<KotCzarny>
you can have fully functional one in <16megs
<ErwinH>
Running at 1008 MHz, starting at 1100mV and dropping the CPU_VDDX in 20mV steps, wait until the core is cooled below a certain temp and start over, stop when the temp reaches a certain temp and start over.
<KotCzarny>
what temp are you targetting?
<ErwinH>
50/80 atm.
<KotCzarny>
and keep in mind its very relative to ambient temp
<ErwinH>
It's a prove of concept now. Once I've setup my cooling tower I can do better tests.
<ErwinH>
This is without any for of cooling.
dh1tw has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
silviop has joined #linux-sunxi
<ErwinH>
How long should I test to see if a certain frequency/voltage combination is stable?
<silviop>
i compile u-boot 2017-rc2 for pine 64 ( make ARCH=arm CROSS_COMPILE=aarch64-linux-gnu-) and i obtain uboot.bin
<silviop>
now i should use dd to put to sd or i need other steps ?
<ErwinH>
For H5 you need quite a few more steps, for Pine64 you need take some steps as well.
<silviop>
there is a guide for mainline u-boot ?
<ErwinH>
Don't know what's the best route? Found quite a few guides, but nothing that produced a working SD card.
<silviop>
from what i understand is necessary to add boot0 functionality
<tkaiser>
ErwinH: In case you really want to do already undervoltage testing I strongly recommend to use this Linpack mentioned yesterday. The hpl call then reports corrupted data before you run into kernel crashes. Saves you a lot of time. And everything should be contained in the archive I linked to yesterday.
<terra854>
tkaiser: "Make sure to stop any graphical UI before running this command to make sure that the existing framebuffer can be properly freed and reassigned."
<terra854>
tkaiser: With DRM, there is no need for that
<silviop>
i came from pc world where you attach monitor and automagically vga/dvi/hdmi autodetect resolution
<tkaiser>
silviop: That's EDID and friends. Works not with the shitty BSP HDMI driver now.
<terra854>
And is EDID available on mainline kernel for AW SoCs?
<tkaiser>
terra854: As MoeIcenowy wrote: With jernej's patch this works now (patch for H3)
JohnDoe_71Rus has joined #linux-sunxi
<silviop>
i feel that mainline is same in all platform
<MoeIcenowy>
the shitty BSP also uses EDID, but it only use EDID to choose the available modes from the pre-defined set
<tkaiser>
u-boot does EDID detection, sets up the framebuffer, kernel uses it. No Mali, no acceleration ;)
<MoeIcenowy>
if your monitor do not support any of the predefined mode, it will fail to initialize the mode
<terra854>
Well, framebuffer is going to be phased out
<MoeIcenowy>
but the mainline patch now use all the EDID info to drive screen
Worf has quit [Quit: Konversation terminated!]
<terra854>
iirc, there is no one maintaining the framebuffer as of 4.10
<apritzel>
tkaiser: If I got this correctly, there are patches for proper DRM support in the kernel, somewhat unrelated to that simplefb support driven by U-Boot
<apritzel>
(for H3, that is)
<tkaiser>
apritzel: I lost track there to be honest. And some days ago I searched for the patches on Jef Moine's site and everything related has been gone. Huh?
silviop has quit [Remote host closed the connection]
victhor has joined #linux-sunxi
Gerwin_J has joined #linux-sunxi
vagrantc has joined #linux-sunxi
<ErwinH>
tkaiser: Thanks for linpack
<ErwinH>
Running tests now, starting at 1008MHz, dropping the voltage per frequency. Once I fail a test I raise the voltage with 10mV and lower the frequency by one step.
<tkaiser>
ErwinH: Just as a reference, would be interesting to get the actual benchmark value when running at 1008 MHz
yann-kaelig has joined #linux-sunxi
<ErwinH>
Sure, I'll run it in a bit.
<ErwinH>
I also record the maximum temperature. (without cooling atm.)
<ErwinH>
912MHz@1040mV = 73C
reinforce has joined #linux-sunxi
<tkaiser>
ErwinH: I would believe the test runs too fast to get reasonable temperature/throttling effects. But of course you should monitor this (and maybe disable throttling for this test?)
<ErwinH>
Throttling? I disabled the entire cpufreq-dt driver and sy8106a driver, and control the frequency / voltage by hand :D
<jrg>
going to install newer Armbian img on my opi+2e some time today if I get a chance.
<tkaiser>
ErwinH: Great! Value looks ok, just compared with my results with Pine64+ back then and interpolated (seems I suffered from some throttling there ;) )
Gerwin_J has quit [Quit: Gerwin_J]
<ErwinH>
Hope my college's will leave soon so I can test at full throttle with proper cooling.
<ErwinH>
cooling is pretty loud with 3 4cm high speed fans.
jstein_ has joined #linux-sunxi
<KotCzarny>
use 12cm or bigger ones
<KotCzarny>
after all only air flow counts
<ErwinH>
If it was something permanent, sure. But this is only for a quick test, so why bother.
<KotCzarny>
for comfort?
jstein_ is now known as jstein
<ErwinH>
If I can't test today, I'll find a room which isn't used. Problem solved ;)
<KotCzarny>
you can also put the board outside the window (assuming you live in northern climate)
<ErwinH>
7C vs 22C, it makes a difference, but not that much...
perr has quit [Remote host closed the connection]
fdcx has quit [Remote host closed the connection]
Andy-D has joined #linux-sunxi
fdcx has joined #linux-sunxi
<KotCzarny>
tkaiser: for distibuting images pixz > zstd. zstd was invented for speed, not compression ratio
<KotCzarny>
and for that it does what's advertised
KB3VGW has joined #linux-sunxi
Amit_T_ has joined #linux-sunxi
KB3VGW has left #linux-sunxi [#linux-sunxi]
<tkaiser>
KotCzarny: It was invented since running code these days for CPUs from 20 years ago is stupid.
<KotCzarny>
you are just jumping on the fad train without considering different use cases
<KotCzarny>
yes, i've read that, and the one on facebook.github.io
<KotCzarny>
and that's what i'm saying to yo. they needed speed, not compression ratio
<KotCzarny>
for distribution makers compression ratio is as important as compression speed though
<KotCzarny>
let's see if i can find any lzham open source implementation
<tkaiser>
KotCzarny: Good to know, maybe I meet someone in the future. I just wanted to explain why zstd can do on a single core where other algorithms need 4. And BTW: I know different use cases and when to choose what.
<tkaiser>
KotCzarny: My point from yesterday was: You missed recent compression algos and you use 'distro packages' and then also a distro that is known for horribly bad results (Debian Jessie)
<KotCzarny>
than it's ok. my use case considers poor uplink when upping packages (only 1mbit)
<KotCzarny>
*then
Amit_T_ has quit [Remote host closed the connection]
<MoeIcenowy>
jernej: I hope you read log and see this sentence. I successfully enabled HDMI and TCON1 on A64, featured a stable full-white display.
<MoeIcenowy>
(as my screen has a simple naive HDMI to RGB chip, the signal stability can be judged from the screen status
<MoeIcenowy>
but it seems that no image is sent to TCON1...
<wens>
is the mux configured?
<MoeIcenowy>
I'm checking...
<MoeIcenowy>
you know, there's no document for DE2
<MoeIcenowy>
wens: is that DE feeds data to TCON, or TCON pull data from DE?
<MoeIcenowy>
oh I mean "who can decide whether to transfer the image data or not, tcon or de"
jernej has joined #linux-sunxi
<MoeIcenowy>
jernej: ping
<MoeIcenowy>
have you see my sentences in log?
atsampson has quit [Ping timeout: 250 seconds]
<NiteHawk>
MoeIcenowy: I'd like to merge https://github.com/linux-sunxi/sunxi-tools/pull/91 soon, which includes (and builds upon) your SID patches. if you find the time, some testing on your devices / feedback would be welcome
matthias_bgg has quit [Quit: Leaving]
robogoat has quit [Ping timeout: 248 seconds]
Amit_T_ has joined #linux-sunxi
<miasma>
KotCzarny: i'd guess zstd would provide nice alternative for squashfs (along with lz4 and xz). it's faster than zlib but provides equal level of compression.
<KotCzarny>
its better than lzo
<KotCzarny>
in this regard it's preferred, but has to be yet implemented in kernel
<miasma>
lzo is already crap. lz4 has similar ratio but better perf
<miasma>
but if you want better ratio than lz4, i think all the current options are suboptimal
<jernej>
MoeIcenowy: I'm extremely limited with time this week.
<jernej>
But I can give you one clue what it could be
<jernej>
just a sec
<MoeIcenowy>
DE is not working at all
<MoeIcenowy>
I'm reading the schematics again to find if there's any power that I misssed
<jernej>
in the sunxi_composer_mode_set() change clrbits_le32(&de_clk_regs->sel_cfg, 1) to setbits_le32(&de_clk_regs->sel_cfg, 1);
<jernej>
afaik, 0 means first tcon and 1 second tcon
<jernej>
but I might be wrong
<MoeIcenowy>
I will check the power supply again
<jernej>
still nothing?
<MoeIcenowy>
as DE is not working at all -- md 01000000 00300000 results in full 0
<jernej>
also the line "writel(1, &de_glb_regs->dbuff);" can be threated as "apply settings"
<jernej>
btw, H5 has additional special setting in HDMI PHY
<jernej>
something about voltage bias
<MoeIcenowy>
I got H5 running now.
<MoeIcenowy>
very easy
<jernej>
working?
<MoeIcenowy>
yes
<MoeIcenowy>
just merge apritzel's code with yours
<MoeIcenowy>
and solve some conflicts in clock driver code
GrimKriegor_ is now known as GrimKriegor
<jernej>
I wonder why is this voltage bias setting there
<jernej>
well, unfortunatelly, I will heavily refactor the code
<jernej>
but then you could easily port your changes over
<jernej>
btw, mailine kernel driver is waiting on Renesas HDMI driver to be merged first in drm-next
<jernej>
if you really want to test DE2, you have to have also some kind of output and HDMI here is easier than TV out
<KotCzarny>
different cheap but technology might be similar
<KotCzarny>
*chip
robogoat has joined #linux-sunxi
<MoeIcenowy>
jernej: mainline has already driver for TV out
<MoeIcenowy>
drivers/gpu/drm/sun4i/sun4i_tv.c
<jernej>
KotCzarny: I saw similar parameters also on other platforms, but I didn't manage to figure why it is really needed
<jernej>
MoeIcenowy: But I guess nobody try to use it on H3. Did you?
<MoeIcenowy>
of course I didn't
<jernej>
so there is no guarantee that it is directly compatible
<jernej>
at least I would go with HDMI first, because I know it well and most users would use it too
<jernej>
sorry, I don't have time anymore for chat
mzki has quit [Ping timeout: 258 seconds]
<jernej>
just write and I will read everything in a day or two
<MoeIcenowy>
ok thx
<jernej>
bye
<MoeIcenowy>
bye
atsampson has joined #linux-sunxi
menomc is now known as mnemoc
<jernej>
MoeIcenowy: It seems that I have few minutes left. Can you please push your A64 WIP branch somewhere? I'm really interested what have you done till now