<ssvb>
BorgCuba: you can also try to compile and run https://github.com/ssvb/lima-memtester/ just to see whether it works and shows a spinning cube (it should support the r3p2 mali kernel driver)
<ssvb>
BorgCuba: with the lima driver, at least everything is open source and you can debug it :-)
<ssvb>
BorgCuba: lima-memtester also includes 'lima-textured-cube' program, which shows animated cube
<ssvb>
BorgCuba: you can run it and see if the cpu is 100% loaded or not
<BorgCuba>
ssvb, when I run lima-memtester I get lots of "Error: failed to run ioctl on /dev/fb0: Invalid argument"
<ssvb>
BorgCuba: does it at least show the cube?
<BorgCuba>
ohh and now that message from the driver appears -> MALI_IOC_MEM_MAP_EXT
<BorgCuba>
yes
<BorgCuba>
its spinning
<BorgCuba>
fullscreen
<BorgCuba>
with a little bit of flickering
<BorgCuba>
htop shows no core is nowhere near 100%
<ssvb>
ok, now it is strange that the mali blob can't do the same
<BorgCuba>
sorry, my mistake: memtester stopped
<ssvb>
why has it stopped?
<ssvb>
it is better to run the 'lima-textured-cube' program, because it is a pure GPU workload
<BorgCuba>
its again one core at 100%
<ssvb>
while the memtester program also uses CPU to test memory
<ssvb>
do you mean that 'lima-textured-cube' causes 100% load for one CPU core?
<BorgCuba>
no, the lima-textured-cube is not on 100%
<BorgCuba>
the core with highest load is at about 12%
levd has joined #linux-rockchip
<BorgCuba>
for textured cube only
<BorgCuba>
does this suggest that the gles implementation is responsible for the 100% workload (on 1 core) during glmark2?
levd1 has joined #linux-rockchip
<ssvb>
BorgCuba: it means that the mali blob somehow can't map the framebuffer memory and fallbacks to using memcpy
<ssvb>
BorgCuba: can you check what kind f ioct was this "Error: failed to run ioctl on /dev/fb0: Invalid argument"?
levd has quit [Ping timeout: 245 seconds]
levd1 is now known as levd
<BorgCuba>
ssvb, must be one of these: FBIOGET_VSCREENINFO, FBIOGET_FSCREENINFO, FBIOPAN_DISPLAY
<BorgCuba>
the error message is displayed regularly (about every sec.) so I would vote for FBIOPAN_DISPLAY
<ssvb>
ok, that's probably fine
levd1 has joined #linux-rockchip
levd has quit [Ping timeout: 245 seconds]
levd1 is now known as levd
<BorgCuba>
oh, now the 4.4-rc6 kernel boots (I own a radxa lite and used the radxa pro dtb (as this ships with the kernel) and forgot to decrease ram size to 1gig)
cnxsoft has joined #linux-rockchip
afaerber has quit [Ping timeout: 260 seconds]
Ueno_Otoko has quit [Ping timeout: 250 seconds]
<BorgCuba>
ssvb, thnaks so far
<BorgCuba>
gotta go
<BorgCuba>
bye
BorgCuba has quit [Ping timeout: 246 seconds]
levd1 has joined #linux-rockchip
levd has quit [Ping timeout: 245 seconds]
levd1 is now known as levd
afaerber has joined #linux-rockchip
levd1 has joined #linux-rockchip
levd has quit [Ping timeout: 255 seconds]
levd1 is now known as levd
Ueno_Otoko has joined #linux-rockchip
hipboi has quit [Read error: Connection reset by peer]
hipboi has joined #linux-rockchip
hipboi has quit [Ping timeout: 245 seconds]
levd1 has joined #linux-rockchip
levd has quit [Ping timeout: 255 seconds]
levd1 is now known as levd
hipboi has joined #linux-rockchip
tlwoerner has quit [Quit: Leaving]
levd1 has joined #linux-rockchip
levd has quit [Ping timeout: 255 seconds]
levd1 is now known as levd
vickycq has joined #linux-rockchip
hipboi has quit [Ping timeout: 240 seconds]
hipboi has joined #linux-rockchip
cyteen_ has quit [Ping timeout: 250 seconds]
levd1 has joined #linux-rockchip
levd has quit [Ping timeout: 255 seconds]
levd1 is now known as levd
amstan_ is now known as amstan
jusched has joined #linux-rockchip
Ueno_Otoko has quit [Ping timeout: 260 seconds]
hipboi has quit [Quit: Leaving]
Ueno_Otoko has joined #linux-rockchip
hipboi has joined #linux-rockchip
<akaizen>
ah the reason that linux-sunxi wiki is so good is libv re-write most of it :D nice job libv!
c0d3z3r0 has quit [Excess Flood]
c0d3z3r0 has joined #linux-rockchip
levd1 has joined #linux-rockchip
levd has quit [Ping timeout: 255 seconds]
levd1 is now known as levd
levd1 has joined #linux-rockchip
levd has quit [Ping timeout: 255 seconds]
levd1 is now known as levd
levd1 has joined #linux-rockchip
levd has quit [Ping timeout: 255 seconds]
levd1 is now known as levd
levd1 has joined #linux-rockchip
levd has quit [Ping timeout: 255 seconds]
levd1 is now known as levd
dowu has joined #linux-rockchip
<dowu>
To make "rootfs.ext4, as I know, I need to do <apt-get install ssh usbutils wpasupplicant wireless-tools>.
<dowu>
But in my place, I could not run this command correctly because a server "daum.net" near my place does not support the preinstalled packages.
<dowu>
So I would get those pre-installed packages from other site. Is it possible? then how can I execute commands?
<aborche>
dowu what you want to do ?
<aborche>
use existing rootfs from radxa
<aborche>
from rabian
<dowu>
During embedding Linux on rk3188, I need to add some packages
<aborche>
hm. existing rootfs already have need packages
<dowu>
I built my own rootfs and it works very fine. But I need to add some more packages which are not included ....
<dowu>
For example, how about emacs?
<aborche>
if you need add some packages, download rootfs.img, mount it with losetup, copy deb files from another repo and install it if you need
<aborche>
or simply add another repo like linaro
<aborche>
and get packages from it
<dowu>
Yes, I need to know how to get other repo.
<aborche>
???
<dowu>
any options on apt-get install command?
<aborche>
in /etc/apt/sourses.list.d/ make a new file with repo info then run apt-get update
<dowu>
to get other repo, e.g. as you said linaro?
<aborche>
yep or raspberry pi
<aborche>
QTFG linaro arm repo
<dowu>
Thank you I will try it. You may give me the answer.
<aborche>
nope :)
<aborche>
hipboi, are you here ?
levd has quit [Ping timeout: 276 seconds]
diego71 has quit [Read error: Connection reset by peer]
mikea has joined #linux-rockchip
mikea is now known as ariellea
naobsd has joined #linux-rockchip
Ueno_Otoko has quit [Ping timeout: 246 seconds]
Ueno_Otoko has joined #linux-rockchip
ganbold has quit [Quit: This computer has gone to sleep]
ganbold has joined #linux-rockchip
ganbold has quit [Remote host closed the connection]
ganbold has joined #linux-rockchip
wsy has joined #linux-rockchip
<aborche>
http://pastebin.com/fGnQRmsZ huh. 4.4-rc6 boot is ok at 3188. but i have some problems with rtc and sdio.
wsy has left #linux-rockchip [#linux-rockchip]
<aborche>
kernel cannot use gpt extended partition as root!
<aborche>
only primary, only dos :(
<naobsd>
aborche: configure your kernel as you want
<naobsd>
aborche: you shouldn't assume mainline/RK kernel/someone's private tree/etc/etc are compatible
<naobsd>
aborche: you shouldn't assume everyone knows well about your device/your work
VargaD has quit [Ping timeout: 260 seconds]
<naobsd>
there are several kernels, mainline(based wip branches), chromeos, RK 3.10/3.0, and someone's private
<naobsd>
device, source tree, and sometimes toolchain need to be explained clearly
<naobsd>
"my device is very similar to XXX" doesn't explain anything
<aborche>
naobsd, hm. if tuned old kernel work is ok, why newest kernel don't work ? i'm not understand. for what too many kernels, branches, device trees if they is not compatible ?
<naobsd>
Rockchip 3.10/3.0 tree is maintaned by Rockchip, ChromeOS kernel tree is maintained by ChromeOS project, mainline kernel is maintained by kernel.org and linux kernel community
<naobsd>
why do you assume they are same?
<aborche>
i can make a list with all peripherals located at board, but i'm sure this is not a problem why rk3188 board which compiled from recommended tech docs from manufacturer not work as need
<naobsd>
you don't assume your RK3188 board is 100% compatible as Radxa Rock, right? you know they are designed differently
<naobsd>
it's same that different kernel from different project are not compatible
<naobsd>
not _100%_ compatible
<naobsd>
some parts are compatible of course
<naobsd>
but not all
<aborche>
i know. and i know some differences in my board and radxa rock/lite board
<naobsd>
same for kernel trees
<aborche>
radxa rock is development board which has been designed for create some open algorimts and programs which be closed by manufacturer into compiled libs, modules and blobs.
<aborche>
this is correct ?
<naobsd>
well, I'm not talking and have no idea about RR characteristics, sorry
<aborche>
ok.
<aborche>
technical progress is not stay. but we have many hardware which not supported after 1 year.
<aborche>
i have mele a3700 based at a10 chip. last firmware update stops at kernel 3.0.8+.
<aborche>
this is not a normal!
<naobsd>
3.0 is old and "no update" is absolutely not good. but sometimes I saw some people "3.x is older than 4.x! too bad"
<naobsd>
some well maintained major Android devices use 3.x too ;)
VargaD has joined #linux-rockchip
<naobsd>
personally I don't expect too much from cheap device
<aborche>
naobsd, also i can say '3.0.36+' work, and 4.x not work. i think this is a manufacturer support problem.
<aborche>
i'm search work kernel which can support rk3188 from 3.10.xx branch, but no result. only some patches from radxa, and again hdmi is not work, old wifi drivers and more more.
<naobsd>
because Rockchip doesn't support 3.10 kernel for RK3188
<naobsd>
I cannot find any reason to support 3.10 kernel by manufacturer
<aborche>
ok. and what we can do with it ?
<naobsd>
develop it, or forget it
<aborche>
hm. you cannot sell a new car and stop make any replacement parts for it after 6 month.
<aborche>
i'm right ?
<naobsd>
why do you think so? you can sell new car and new replacement parts as long as you can make them
ckeepax has quit [Ping timeout: 276 seconds]
<aborche>
but in this situation you not receive any new replacement parts like software patches from manufacturer
<aborche>
and you need drop current hardware to trash and buy a new, which not be supported after some time
<naobsd>
manufacturer promised they make replacement parts as long as you want? you paid for such a service?
<aborche>
i'm ready to pay for service.
<naobsd>
so if you can find good manufacturer who gives good update service, you'll be happy
<aborche>
yep. but i'm sure i cannot find it.
<naobsd>
I don't know such a manufacturer (I'm not business person in this area)
<aborche>
me too. as sysadm i know what my hardware can work very long time. i have servers with 5 years uptime and i can update system and it be work as need
<aborche>
ok. this is very long dialog. sorry.
nighty^ has quit [Remote host closed the connection]
gb_master has quit [Remote host closed the connection]
case1 is now known as xcasex
jas-hacks has joined #linux-rockchip
<BorgCuba>
an idea why the 4.4-rc6 kernel for rk3188 is so slow?
<ariellea>
mmind00: 100ms works for my panel in veyron-speedy too
<mmind00>
ariellea: thanks for testing ... then we'll go with that until someone with a slower (i.e. > 100ms) panel comes along
<ariellea>
mmind00: yup, sounds good :)
<BorgCuba>
mmind00, do you have any clue why the rk3188 performs significantly worse on a mainline kernel (4.4-rc6 in this case)?
<BorgCuba>
compared to 3.0.36+ rk custom kernel
<mmind00>
BorgCuba: not sure what you mean with "performs worse", if you mean feature-wise, that's simply because currently nobody cares enough about the older platforms to invest a lot of time in them
<BorgCuba>
now, I mean benchmark wise
<BorgCuba>
roughly between 2 to 3times slower in nbench
<BorgCuba>
dhrystone 0.8ms (4.4) 0.3ms (3.0)
<mmind00>
BorgCuba: ddr-speed maybe ... we currently don't have any ddr-frequency scaling, and uboot starts the ram in some save speed if I remember correctly
<BorgCuba>
I think it is clocked at 1.2GHz instead of 1.6 but still
<BorgCuba>
mmind00, can I find these infos in /sys or /proc ?
<mmind00>
mount none /somewhere -t debugfs; cat /somewhere/clk/clk_summary ... on mainline
<mmind00>
BorgCuba: for the 3.0-stuff you will need to find some sort of similar output yourself
<BorgCuba>
okay, I'll check
<BorgCuba>
thnaks
<BorgCuba>
mmind00, the highest rate I found was 768e6. Maybe you could take a look and tell me which clock is for the cpus and which one form the dram? -> http://pastebin.com/Sqx3dY95
<mmind00>
BorgCuba: "ddrphy" running at 300MHz is the ram
<BorgCuba>
alright, thank you
<mmind00>
BorgCuba: and your "armclk" seems to also run at 600MHz
<mmind00>
which is the boot-default
<BorgCuba>
indeed
<BorgCuba>
so the results make sense I guess
<mmind00>
cpufreq is working, after you define the supplying regulator
<mmind00>
BorgCuba: yep, that is the needed assignment ... if you are on a rock, you'll need to make sure the act8865/act8846 driver + cpufreq-dt is compiled in
<BorgCuba>
let me see
<BorgCuba>
from .config: CONFIG_REGULATOR_ACT8865=y
<BorgCuba>
and a "act8846" node in the dts: "compatible = "active-semi,act8846";"
<BorgCuba>
but no cpufreq
<mmind00>
BorgCuba: you are on a radxarock?
<BorgCuba>
yes, lite version
<BorgCuba>
I already adjusted the memory size to 1gig
<mmind00>
CONFIG_CPUFREQ_DT=y is also set?
<BorgCuba>
no
<BorgCuba>
I will activate it
<mmind00>
it should be :-)
<BorgCuba>
so I activated "Generic DT based cpufreq driver"
<BorgCuba>
rebuilding ...
<BorgCuba>
so I will try to run nbench again with the new kernel