Ubuntu U2 projects status DASHBOARD

Moderators: odroid, mdrjr

Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Fri May 31, 2013 10:07 am

I am creating this thread as a reference for myself and other newbies to ODROID. It's supposed to keep track of different developments in projects that I am interested in, as a reference for other people as well.

Last Updated: 05/09/2013 - XBMC instructions
Updated: 26/08/2013 - new results for 3.8.13 after mali 3D fix
Updated: 09/06/2013 - reinstalled Xubuntu from hardkernel image
Updated: 07/06/2013 - switched to Debian (720p), 3D fixed, insanely fast in 3.0.80 (>600fps in es2gears!), 3.8.13 fixes:reboot and no more 3D tearing; acceleration still not as good (~half the framerate from 3.0.80) but CPU utilization seems smaller
Updated: 01/06/2013 - Ubuntu initial results

3.0 kernel dmesg example: http://paste.debian.net/30677/
3.8 kernel dmesg example: http://paste.debian.net/33435/

OS: Ubuntu 13.04 installed from the image provided at http://forum.odroid.com/viewtopic.php?f=8&t=1193
Packages: removed bloat, installed dev packages for compiling retroarch, xbmc; list of packages currently installed: http://pastebin.com/gpeJ7qUD
OS: Debian Wheezy installed from retroach image provided by neagix; list of packages currently installed: [url][/url]

Kernels:
  • 3.0.80 compiled using the image provided config, with mali drivers installed, output on HDMI 1080p TV
  • 3.8.13 cloned git repository from https://github.com/hardkernel/linux/tree/odroid-3.8.y, pull and compile from time to time, howto at http://forum.odroid.com/viewtopic.php?f=52&t=81
    • FIXED! using the raspberry pi for UART! (see post below) reboot not working (can't debug, I have no UART board) - apparently it's an issue with 3.6.y+ kernels in arch/arm/mach-exynos/cpuidle.c (see post below)
    • mali acceleration not working properly (roughly half the FPS from 3.0.80 with more less CPU usage) - FIXED
    • uInitrd from default defconf uncompressed at >20MB results in hang. Using my own kernel defconf with less modules. uInitrd not working, have to use the uInitrd from 3.0.75
    • output on HDMI 1080p TV using new kernel options hdtv_type=hdmi hdtv_format=1080p50hz (removed the kernel config option CONFIG_ODORID_HDMI_SW_CONFIG + small patch on the sources from 20/06/2013 http://pastebin.com/Jd4HXwAV)

Mali acceleration
  • /etc/X11/xorg.conf at http://pastebin.com/QCYFYad2
  • 3.0.80
    • glxgears: software renderer, ~120fps with 1 core @ 100%
    • es2gears: EGL_VERSION: 1.4 Linux-r3p2-01rel0, 560-600+fps with ~ 1 core @ 40%, 1 core @ 40-100%
    • glmark2-es2: GL_RENDERER: Mali-400 MP, GL_VERSION: OpenGL ES 2.0, ~ 1 core @ 50-90%, 2 cores @ <20%, score: 140fps Debian, 137fps Ubuntu (fullscreen 80 fps Debian, 73 Ubuntu)
  • 3.8.13
    • glxgears: software renderer, ~ 120fps with 1 core @ 100%
    • es2gears: EGL_VERSION: 1.4 Linux-r3p2-01rel0, 560-580fps with ~ 2 cores @ 30% ; 635-665fps overclocked to 640MHz
    • glmark2-es2: GL_RENDERER: Mali-400 MP, GL_VERSION: OpenGL ES 2.0, ~ 1 core @ 40% (up to 80%, +20% core2 for some tests), score: 143fps Debian (81 fullscreen) - a bit better when overclocked at 640Mhz: 148fps, 82fps.

X11 drivers: compile instructions at: viewtopic.php?f=15&t=2576&p=20685#p20685

RetroArch: compiled according to http://neagix.blogspot.com.au/2013/04/retroarch-on-odroid-u2-with-hardware.html
  • running in window mode SNES9x: 50fps on 3.8.13, 50fps on 3.0.80, 1 core @50%, another core with low utilization, working very nice
  • running in full screen mode - same as above
  • 3.8.13: without the regulator fix, there would be visual effects not rendered properly; it's been fixed now
  • ~/.retroarch.cfg at http://pastebin.com/9AW50Psv

XBMC 13:
  • source from https://github.com/mdrjr/xbmc/commits/master (use Frodo branch for 12.2)
  • config options:
    Code: Select all
    ./configure --disable-pulse --enable-alsa --disable-libcec --enable-neon --disable-debug --enable-optimizations --enable-ccache --enable-gles --enable-non-free --enable-exynos4 --disable-gl --disable-vdpau --disable-vaapi --disable-crystalhd --disable-openmax --disable-joystick --disable-rsxs --disable-projectm --disable-fishbmc --disable-nfs --disable-afpclient --disable-dvdcss --disable-optical-drive --disable-libbluray --enable-texturepacker --host=armv7a-hardfloat-linux-gnueabi --with-platform=samsung-exynos --with-cpu=cortex-a9

  • working binary, accelerated GUI (GLES), running at 50fps (excellent) on 3.0.80 and 70fps on 3.8.13 (640MHz GPU overclock). Using mfc hardware acceleration to play videos. Crashes on exit.
  • XBMC (&ffmpeg) compiled with:
    Code: Select all
    FFMPEG_CFLAGS="-march=native -mfloat-abi=hard -mfpu=neon -ftree-vectorize -mvectorize-with-neon-quad -ffast-math -mcpu=cortex-a9 -mtune=cortex-9 -O3 -pipe -fstack-protector --param=ssp-buffer-size=4 -D_FORTIFY_SOURCE=2" CFLAGS="-march=native -mfloat-abi=hard -mfpu=neon -ftree-vectorize -mvectorize-with-neon-quad -ffast-math -mcpu=cortex-a9 -mtune=cortex-a9 -O3 -pipe -fstack-protector --param=ssp-buffer-size=4 -D_FORTIFY_SOURCE=2" CXXFLAGS="-march=native -mfloat-abi=hard -mfpu=neon -ftree-vectorize -mvectorize-with-neon-quad -ffast-math -mcpu=cortex-a9 -mtune=cortex-a9 -O3 -pipe -fstack-protector --param=ssp-buffer-size=4 -D_FORTIFY_SOURCE=2" make -j5

  • XBMC 13 seems to work better than 12.2 (lower CPU usage)
  • need to have pulse audio installed and started (pulseaudio --start), even if XBMC is compiled with ALSA support it will use the pulse driver that falls back on ALSA
  • /etc/asound.conf sound configuration:
    Code: Select all
    pcm.!default {
      type pulse
    }
    ctl.!default {
      type pulse
    }

  • required 3.8 kernel, with CMA memory 128MB and MFC memory 64MB for 1080p decoding
  • use with Debian-based distributions, Ubuntu X11 has half the EGL performance
  • tested with both mali and sunxifb, mali performs better, especially on 1080p resolution
  • on 720p resolution: MFC decoder able to play any video (720p, 1080p) without drops
  • on 1080p resolution: MFC decoder able to play 720 video without drops; 3-4 drops/sec on 1080p video and one core maxed at 100% CPU, with lots of kernel usage - looks like there is room for optimizations.
  • --------------------
  • other (old) sources: https://github.com/koying/xbmc/commits/odroid, https://github.com/huceke/xbmc/tree/exynos
Last edited by memeka on Tue Nov 19, 2013 5:44 pm, edited 22 times in total.
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby odroid » Fri May 31, 2013 10:29 am

Changed to STICKY because this post is really essential for newbies as well as maintainers. ;)
User avatar
odroid
Site Admin
 
Posts: 22276
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: Ubuntu U2 projects status DASHBOARD

Unread postby OverSun » Fri May 31, 2013 12:48 pm

uInitrd is working on 3.8
reboot is working on 3.8
HDMI output with and without the option is a feature, not an issue, and anyway yesterday it was reworked completely in git.
User avatar
OverSun
 
Posts: 1186
Joined: Mon Apr 29, 2013 5:12 pm
languages_spoken: english

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Fri May 31, 2013 1:03 pm

OvverSun wrote:uInitrd is working on 3.8
reboot is working on 3.8
HDMI output with and without the option is a feature, not an issue, and anyway yesterday it was reworked completely in git.


I have compiled yesterday's version.
In my case reboot leads to a shutdown, damn that UART console that I don't have. Maybe it's because of the old uInitrd.
I have generated the uInitrd like this:
Code: Select all
update-initramfs -c -k 3.8.13+
mkimage -A arm -O linux -T ramdisk -C none -a 0 -e 0 -n uInitrd -d /boot/initrd.img-3.8.13+ /media/boot/uInitrd

and with it the U2 won't boot (again, damn that UART)
Any idea what I'm doing wrong? (I've seen the uInitrd is also missing from mdrjr's builder images).

EDIT: my kernel would not compile when removing the HDMI option, I had to change a bit a file
EDIT: just saw today's "eMMC: fixing reboot lockup while using eMMC" commit, rebuilding again...
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby OverSun » Fri May 31, 2013 3:49 pm

There is no need to mess with initrd generation by yourself
Add '-odroidu2' as an external version to kernel, edit a scripts/setlocalversion to remove '+' at the end, so your kernel name will have a name '3.8.13-odroidu2' and do a 'make zinstall', the process will generate and place on /media/boot everything for you automatically.
I had sometimes nonbooting kernels, I've found out two reasons for this:
1. Too quick 'reboot' after kernel installation, obviously do a 'sync' before reboot would be a great idea. Looks like cache was not always flushed to fat32 partition on reboot.
2. Too big initrd file because of huge amount of modules, it boots from time to time. What I did is just enabled 'Use kernel /proc.config.gz' in kernel, boot, and compile a kernel with 'make loclayesversion', so you will have the kernel configured with 'YES' on all options you currently have in kernel plus 'YES' for all the modules currently loaded in kernel. That produces initrd with no modules (because none needed), and this one boots 100% every single time.

PS. I too don't have an UART board, I've just connected my ODROID to my Raspberry Pi, because it has 1.8v UART and it fits just perfect. =) minicom on rPI to com1 and you have a console! ;)
User avatar
OverSun
 
Posts: 1186
Joined: Mon Apr 29, 2013 5:12 pm
languages_spoken: english

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Fri May 31, 2013 3:54 pm

Thanks for the tips, especially about raspberry pi :)
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby OverSun » Fri May 31, 2013 4:08 pm

Untitled.jpg


Then you will have to do the procedure described https://sites.google.com/site/semilleroadt/raspberry-pi-tutorials/gpio at "UART" section.

1. Disable raspberry console on com1 (to free the port)
2. Reboot
2. Start a minicom on /dev/ttyAMA0
User avatar
OverSun
 
Posts: 1186
Joined: Mon Apr 29, 2013 5:12 pm
languages_spoken: english

Re: Ubuntu U2 projects status DASHBOARD

Unread postby recliq » Fri May 31, 2013 6:30 pm

I think this is setup is a bit dangerous because AFAIK RPi has 3.3V UART/GPIO whereas Odroid has 1.8V UART!
Maybe you should place a level shifter in between the two, eg. like this one:
https://www.sparkfun.com/products/8745
User avatar
recliq
 
Posts: 110
Joined: Mon Feb 25, 2013 6:07 pm
Location: Germany
languages_spoken: english, german
ODROIDs: 2xU2 + XU3 + XU4

Re: Ubuntu U2 projects status DASHBOARD

Unread postby sert00 » Fri May 31, 2013 9:40 pm

i think recliq is right,and a level shifter is so inexpensive that is better to put it in-between
I have the HK UART,but the tip of ovversun is very great,and i want to try myself with a level-shifter\converter
2x Odroid-U2
2x Odroid-XU-E
Odroid-U3
Odroid-C1+
Odroid-XU4
Lcd touch panel 10.1"
....and many of the goodies available....
User avatar
sert00
 
Posts: 745
Joined: Sun Feb 24, 2013 12:26 am
languages_spoken: english,italian
ODROIDs: ODROID-U2*2__ODROID-U3__ODROID-XU-E__ODROID-XU-Ebeta__ODROID-C1__ODROID-XU4

Re: Ubuntu U2 projects status DASHBOARD

Unread postby mdrjr » Fri May 31, 2013 10:01 pm

Quick toughs...

uInitrd does work, however the issue that you have is that your uInitrd is exceeding 20Mb that is a limitation of the bootloader, not a kernel issue. I've disabled a few modules to attempt to lower the size of it. However I've not tested the new size yet, since I don't use uInitrd. I'll probably look into the size of it..
uInitrd isn't needed on ODROID, since all the needed hardware to boot the board is already as built-in on the kernel.

On reboot:
There was two issues. One is a kernel panic when you reboot. This issue still persists. (This issue is present for all Exynos board and as far as I recall it existed on kernel 3.6 too). Its something on arch/arm/mach-exynos/cpuidle.c

Another issue was that when you used the eMMC module and the bug above didn't occur it would simple lock up. That was my fault because it lacked the reset code of the eMMC. This issue has been fixed.

HDMI:
I've added several new options including DVI or HDMI mode and alot of new resolutions supported. Please check the latest commit as usual and refer to the tools/hardkernel folder inside the kernel. It has all the boot.scrs for our Ubuntu 13.04 with all possible options.

Mali:
I'm doing some tests on this issue and all I can say it that right now, it doesnt appear to be a issue on Mali itself but on the kernel. Mali and UMP relays heavily on Kernel Memory to work and for some unknown reason this kind of memory access is slow on 3.8. Not sure why..

Apart from those two bugs mentioned... everything its working fine.
mdrjr
Site Admin
 
Posts: 11581
Joined: Fri Feb 22, 2013 11:34 pm
Location: Brazil
languages_spoken: english, portuguese
ODROIDs: -

Re: Ubuntu U2 projects status DASHBOARD

Unread postby neagix » Sat Jun 01, 2013 4:33 am

memeka wrote:OS: Ubuntu 13.04 installed from the image provided at http://forum.odroid.com/viewtopic.php?f=8&t=1193
Packages: removed bloat, installed dev packages for compiling retroarch, xbmc; list of packages currently installed: http://pastebin.com/gpeJ7qUD

That's why I did a minimal Debian image, to not have to deal with bloat ;)
memeka wrote:Mali acceleration
  • /etc/X11/xorg.conf at http://pastebin.com/QCYFYad2
  • 3.0.75
    • glxgears: software renderer, ~120fps with 1 core @ 100%
    • es2gears: EGL_VERSION: 1.4 Linux-r3p2-01rel0, 100-160fps with ~ 1 core @ 40%, 1 core @ 40-100%
    • glmark2-es2: GL_RENDERER: Mali-400 MP, GL_VERSION: OpenGL ES 2.0, ~ 1 core @ 50-80%, 2 cores @ 20%, score: 85-95fps

The glmark2-es2 score is slow! Please check my hardware acceleration tests.

memeka wrote:RetroArch: compiled according to http://neagix.blogspot.com.au/2013/04/retroarch-on-odroid-u2-with-hardware.html
  • running in window mode SNES9x: 30-35fps on 3.8.13, choppy sound; 40-45fps on 3.0.75, sound acceptable but not perfect, overall I think running worse than pocketsnes on raspberry pi
  • running in full screen mode ???
  • ~/.retroarch.cfg at http://pastebin.com/9AW50Psv

On 3.8, it's currently impossible to get it fast. But for 3.0.75 I strongly doubt you did it right.

I have perfect sound and maximum FPS rate. I could achieve it also on Ubuntu, so it's either your desktop environment or wrong Mali drivers configuration; just wanted to add the precisation so that you know it's definitively possible to get it right ;)
neagix
 
Posts: 264
Joined: Sun Apr 14, 2013 2:54 am
languages_spoken: english
ODROIDs: U2

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Sat Jun 08, 2013 8:24 am

updated!
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby neagix » Sat Jun 08, 2013 3:40 pm

memeka wrote:Last Updated: 07/06/2013 - switched to Debian (720p), 3D fixed, insanely fast in 3.0.80 (>600fps in es2gears!)

Glad you've tried my bare-metal distro :) but...wait a minute. >600? Seriously? I have measured about 345 FPS so far!

Please tell me more :P
neagix
 
Posts: 264
Joined: Sun Apr 14, 2013 2:54 am
languages_spoken: english
ODROIDs: U2

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Sat Jun 08, 2013 4:32 pm

Yeah, insanely fast :)
Did a make headers install and it deleted most of them :( need to start from scratch now :(
I'll retry on Ubuntu first and see if I can get better results, my impression is that X server influences things greatly...
I was using 3.0.80 kernel and the updated Xorg server...
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby neagix » Sat Jun 08, 2013 4:46 pm

memeka wrote:Yeah, insanely fast :)
Did a make headers install and it deleted most of them :( need to start from scratch now :(
I'll retry on Ubuntu first and see if I can get better results, my impression is that X server influences things greatly...
I was using 3.0.80 kernel and the updated Xorg server...

Yes, it's the X server that makes the difference.

But 600? I don't think that's possible :)
neagix
 
Posts: 264
Joined: Sun Apr 14, 2013 2:54 am
languages_spoken: english
ODROIDs: U2

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Sat Jun 08, 2013 6:43 pm

it was 602 :)
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby neagix » Sat Jun 08, 2013 6:48 pm

memeka wrote:it was 602 :)

Ok, please tell me which version of Xorg and the xorg.conf you used - I want to reproduce this

Or maybe your GPU is overclocked? If yes, how?
neagix
 
Posts: 264
Joined: Sun Apr 14, 2013 2:54 am
languages_spoken: english
ODROIDs: U2

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Sat Jun 08, 2013 7:10 pm

Xorg and xorg.conf installed in your debian image, I will provide the list of extra packages installed (I had it saved a while ago).
I had a custom 3.0.80 kernel configured and the latest xorg mali driver compiled from the arm homepage.
Like I said, I erased the headers by mistake to I am reinstall ubuntu now, I'm hoping to reproduce this in XUbuntu too :D
It was all in the base window manager (lxde is it?) from the image.
GPU was not overclocked in 3.0.80 but it was overlocked in 3.8.13 (though 3.8.13 is still slow, it does not have visual artefacts/tearing anymore). I left a message for mdrjr how to patch the 3.8 kernel...
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby neagix » Sat Jun 08, 2013 7:11 pm

memeka wrote:Xorg and xorg.conf installed in your debian image, I will provide the list of extra packages installed (I had it saved a while ago).
I had a custom 3.0.80 kernel configured and the latest xorg mali driver compiled from the arm homepage.
Like I said, I erased the headers by mistake to I am reinstall ubuntu now, I'm hoping to reproduce this in XUbuntu too :D
It was all in the base window manager (lxde is it?) from the image.
GPU was not overclocked in 3.0.80 but it was overlocked in 3.8.13 (though 3.8.13 is still slow, it does not have visual artefacts/tearing anymore). I left a message for mdrjr how to patch the 3.8 kernel...

Problem solved: now I am testing and I see 600 fps as well with 720p resolution, sorry for the noise - and thanks for letting me find this!
I am also checking how 3.8 is doing

Update: more recent tests here
neagix
 
Posts: 264
Joined: Sun Apr 14, 2013 2:54 am
languages_spoken: english
ODROIDs: U2

Re: Ubuntu U2 projects status DASHBOARD

Unread postby neagix » Sun Jun 16, 2013 11:50 pm

@memeka: did you ever merge to XBMC the patches you mentioned in OP?

To make the point about XBMC:

  • no MFC or G2D acceleration is working right now, neither on 3.0 or 3.8
  • XBMC development is focusing energies on 3.8 mostly
  • the only usable XBMC is the one for 3.0 (See this topic)

Confirm/deny?
neagix
 
Posts: 264
Joined: Sun Apr 14, 2013 2:54 am
languages_spoken: english
ODROIDs: U2

Re: Ubuntu U2 projects status DASHBOARD

Unread postby OverSun » Tue Aug 06, 2013 5:09 am

Sorry to interrupt you guys, but could you clarify for me how to overclock mali on 3.8 kernel. I badly want to test the tearing gone and performance things.
Also maybe someone of you would be able to figure out how we can overclock MFC or even better FIMC from their 200Mhz a little higher. It was possible on 3.0, but there are no such clock setup for these devices in 3.8 anymore.
User avatar
OverSun
 
Posts: 1186
Joined: Mon Apr 29, 2013 5:12 pm
languages_spoken: english

Re: Ubuntu U2 projects status DASHBOARD

Unread postby Matt » Tue Aug 06, 2013 1:36 pm

The 3.8.y kernel doesn't have any overclock options in the configuration, but you can do it by editing the kernel source code. If you need my help, I'd be willing to attempt to track down where the default GPU frequency is set when Mali DVFS isn't enabled.
ODROID Boards: U2, X2, XU+E rev.2
Accessories: VU, Smart Power, 13" LCD Kit (for the X2)
User avatar
Matt
 
Posts: 576
Joined: Tue Feb 26, 2013 4:57 pm
Location: Charlottesville, VA, USA
languages_spoken: English
ODROIDs: X2, U2, XU+E rev.2, Smart Power

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Tue Aug 06, 2013 1:46 pm

I don't have my linux image on my emmc, but these where the changes I did:

* add the platform file in Kbuild (currently is not compiled)
https://github.com/hardkernel/linux/blob/odroid-3.8.y/drivers/gpu/arm/mali/Kbuild change:
mali-$(CONFIG_MALI_DVFS) += platform/pegasus-m400/mali_platform.o
to:
mali-y += platform/pegasus-m400/mali_platform.o

* change the voltage and clock from platform file (see the source code from 3.0 for good values)
https://github.com/hardkernel/linux/blob/odroid-3.8.y/drivers/gpu/arm/mali/platform/pegasus-m400/mali_platform.c lines 66-67

I'm not sure if there is anything else... Should have saved the diffs before Android...

You should get "regulator_enable" message in dmesg if platform file is being used, else you need to make sure is actually compiled. Not sure if you need to add arm.o in Kbuild as well...


EDIT: from my experience, there is no visible increase in performance (I have tried 440MHz vs. 640MHz) because it's not the actual GPU performance that limits the performance, but an increase amount of system calls, so the system spends huge amount of time just switching kernelmode-usermode, and less amount doing actual work. I have spent lots of time debugging why is that, with no success :( I would suggest instead:
* try Lima driver (dunno what's the current status)
* try newer driver and newer libMali (provided hardkernel has the newer libMali code and can find time for it, given the XU developments)
* somebody else try to debug why there are 2x more GP job ioctls in the 3.8 kernel
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby OverSun » Tue Aug 06, 2013 3:07 pm

Thanks for the tip about platform file, I'll try to use it.
And regarding everything else, yes, I'm pretty aware of what is the situation.
Preferably myself I would go newer libMali.so way. The problem is that this should be created by Hardkernel, because they have their own blob, which Mali manufacturer gives only to platform creators.
User avatar
OverSun
 
Posts: 1186
Joined: Mon Apr 29, 2013 5:12 pm
languages_spoken: english

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Tue Aug 06, 2013 3:23 pm

OvverSun wrote:Preferably myself I would go newer libMali.so way, the problem is that this should be compiled by Hardkernel because they have their own blob which Mali manufacturer gives only to platform creators.


Exactly my point above :)
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby odroid » Tue Aug 06, 2013 8:08 pm

memeka wrote:
OvverSun wrote:Preferably myself I would go newer libMali.so way, the problem is that this should be compiled by Hardkernel because they have their own blob which Mali manufacturer gives only to platform creators.


Exactly my point above :)


We don't have any update notice from ARM Mali division yet.
As far as I heard from ARM, they have a plan to release new version R4P0 in Q4 this year while we are still using R3P2 now.
I'll try to contact them to get the date more precisely. But the reply need many days as usual.
User avatar
odroid
Site Admin
 
Posts: 22276
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: Ubuntu U2 projects status DASHBOARD

Unread postby OverSun » Wed Aug 07, 2013 1:33 am

I've changed clock rate for xusbxti timer in arch/arm/mach-exynos/mach-hkdk4412.c to 3200000 from 24000000
so according to this line it is now
[ 0.000000] sclk_fimc: source is xusbxti (1), rate is 20000000
20Mhz?
was 15000000
And immediately got 25 fps for 720p and around 20fps smooth for 1080p.
But with that I got garbled output instead of symbols on ttySAC1, so looks like the timer is a source for several devices. =D

Could we consider adding separate timer for FIMC m2m and clock it higher than xusbxti for other devices? As I remember in 3.0 fimc had their own clock rate labeled sclk_fimc as now only mfc has their distinguished clocks sclk_mfc0

Ah, and I see that xusbxti is something like a base to other timers, because all other clocks are higher with it.
And believe me or not, I got 30+ fps in XBMC where it was 15 before.

And you know what? Rising the mali Hz doesn't make a difference to non raised! 440Mhz = 640Mhz in fps, and equals 15 approx.
But rising xusbxti timer made it 30! So there is something wrong with clocking. And clocking is not right for FIMC and Mali for sure.
User avatar
OverSun
 
Posts: 1186
Joined: Mon Apr 29, 2013 5:12 pm
languages_spoken: english

Re: Ubuntu U2 projects status DASHBOARD

Unread postby mdrjr » Wed Aug 07, 2013 3:22 am

but xusbxti must be 24Mhz.
mdrjr
Site Admin
 
Posts: 11581
Joined: Fri Feb 22, 2013 11:34 pm
Location: Brazil
languages_spoken: english, portuguese
ODROIDs: -

Re: Ubuntu U2 projects status DASHBOARD

Unread postby OverSun » Wed Aug 07, 2013 3:28 am

Yeah, I already got it, changing xusbxti skews a lot of things.
But is there possible way to make separate clock for fimc and scale them up a bit? For example clocks for mfc sclk_mfc is using mount_mfc0 and is 35Mhz
[ 0.000000] sclk_mfc: source is mout_mfc0 (0), rate is 55000000
User avatar
OverSun
 
Posts: 1186
Joined: Mon Apr 29, 2013 5:12 pm
languages_spoken: english

Re: Ubuntu U2 projects status DASHBOARD

Unread postby mdrjr » Wed Aug 07, 2013 3:32 am

nope.. mfc frequency is also wrong! MFC should be at 266Mhz as I recall.
mdrjr
Site Admin
 
Posts: 11581
Joined: Fri Feb 22, 2013 11:34 pm
Location: Brazil
languages_spoken: english, portuguese
ODROIDs: -

Re: Ubuntu U2 projects status DASHBOARD

Unread postby OverSun » Wed Aug 07, 2013 3:47 am

The funny thing is that MFC processing is enough for 1080p. It can easily do 30 fps, but FIMC is damn slow and can do only 15 fps max for 1080. But if you scale clock a bit...

Plus on 3.0 FIMC and MFC are tied to same clock with same Hz, and on 3.8 MFC has 4 times faster clock than FIMC
User avatar
OverSun
 
Posts: 1186
Joined: Mon Apr 29, 2013 5:12 pm
languages_spoken: english

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Wed Aug 07, 2013 5:55 am

Wow that's a good find Ovversun
Though I don't get why fimc clock would influence egl fps
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby OverSun » Wed Aug 07, 2013 6:03 am

memeka wrote:Wow that's a good find Ovversun
Though I don't get why fimc clock would influence egl fps


Cause that xusbxti clock is somehow a base for all other clocks. I was firstly looking at fimc and mfc clocks and how they changed and didn't notice, but after I've found out changes in egl framerate, it tuns out that all other clocks "sclk_" in dmesg are scaled up.
Probably egl performance is dependent on some clocking in hdmi output and was bumped because of that.
But obviously rising all the clocks along the device is a bad idea, garbled UART output is less of the issues we could encounter.
I've noticed that sound in XMBC is _slower_ than it should be, and obviously would find out a lot more, so I reverted clocks back and now trying to hook fimc to other clocks than xusbxti, and already successfully did it to mout_mfc1, but for some reason now it is not scaling up properly... I'm trying to figure out where Hz are set for mout_mfc0 and mout_mfc1
User avatar
OverSun
 
Posts: 1186
Joined: Mon Apr 29, 2013 5:12 pm
languages_spoken: english

Re: Ubuntu U2 projects status DASHBOARD

Unread postby neagix » Wed Aug 07, 2013 6:20 am

OvverSun wrote:
memeka wrote:Wow that's a good find Ovversun
Though I don't get why fimc clock would influence egl fps


Cause that xusbxti clock is somehow a base for all other clocks. I was firstly looking at fimc and mfc clocks and how they changed and didn't notice, but after I've found out changes in egl framerate, it tuns out that all other clocks "sclk_" in dmesg are scaled up.
Probably egl performance is dependent on some clocking in hdmi output and was bumped because of that.
But obviously rising all the clocks along the device is a bad idea, garbled UART output is less of the issues we could encounter.
I've noticed that sound in XMBC is _slower_ than it should be, and obviously would find out a lot more, so I reverted clocks back and now trying to hook fimc to other clocks than xusbxti, and already successfully did it to mout_mfc1, but for some reason now it is not scaling up properly... I'm trying to figure out where Hz are set for mout_mfc0 and mout_mfc1


nice finding, OvverSun! as I understand this is also related to VSYNC and the EGL performance delta in 3.8. Actually, it explains it pretty well..
neagix
 
Posts: 264
Joined: Sun Apr 14, 2013 2:54 am
languages_spoken: english
ODROIDs: U2

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Wed Aug 07, 2013 6:56 am

This means that:

- xbmc won't get performance boost from fimc clock
- increasing fimc clock does not mean smooth 1080p, there might be other factors that made it smooth with all clocks higher

The performance issue is still there, xbmc should run at 60fps
But yeah, since mfc and fimc has same clock on 3.0 seems like the scum is platform files are quite different between kernels
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby mdrjr » Wed Aug 07, 2013 7:01 am

memeka wrote:This means that:

- xbmc won't get performance boost from fimc clock
- increasing fimc clock does not mean smooth 1080p, there might be other factors that made it smooth with all clocks higher

The performance issue is still there, xbmc should run at 60fps
But yeah, since mfc and fimc has same clock on 3.0 seems like the scum is platform files are quite different between kernels


does it runs at 60fps on Android @ 1080p?
mdrjr
Site Admin
 
Posts: 11581
Joined: Fri Feb 22, 2013 11:34 pm
Location: Brazil
languages_spoken: english, portuguese
ODROIDs: -

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Wed Aug 07, 2013 7:13 am

Yes
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Wed Aug 07, 2013 10:03 am

mdrjr wrote:nope.. mfc frequency is also wrong! MFC should be at 266Mhz as I recall.


From 3.0.51 kernel (Android):

Code: Select all
[    0.000000] exynos4_setup_clocks: registering clocks
[    0.000000] exynos4_setup_clocks: xtal is 24000000
[    0.000000] uclk1: source is mout_mpll_user (6), rate is 110000000
[    0.000000] uclk1: source is mout_mpll_user (6), rate is 110000000
[    0.000000] uclk1: source is mout_mpll_user (6), rate is 110000000
[    0.000000] uclk1: source is mout_mpll_user (6), rate is 110000000
[    0.000000] sclk_csis: source is xusbxti (1), rate is 1500000
[    0.000000] sclk_csis: source is xusbxti (1), rate is 1500000
[    0.000000] sclk_cam0: source is xusbxti (1), rate is 1500000
[    0.000000] sclk_cam1: source is xusbxti (1), rate is 1500000
[    0.000000] sclk_fimc: source is xusbxti (1), rate is 1500000
[    0.000000] sclk_fimc: source is xusbxti (1), rate is 1500000
[    0.000000] sclk_fimc: source is xusbxti (1), rate is 1500000
[    0.000000] sclk_fimc: source is xusbxti (1), rate is 1500000
[    0.000000] sclk_fimd: source is mout_mpll_user (6), rate is 55000000
[    0.000000] sclk_fimd: source is xusbxti (1), rate is 1500000
[    0.000000] sclk_mfc: source is mout_mfc0 (0), rate is 55000000
[    0.000000] sclk_g3d: source is mout_g3d0 (0), rate is 55000000
[    0.000000] sclk_pwi: source is xusbxti (1), rate is 1500000
[    0.130000] s3c24xx-pwm s3c24xx-pwm.1: tin at 110000000, tdiv at 110000000, tin=divclk, base 8
[    0.130000] s3c24xx-pwm s3c24xx-pwm.0: tin at 110000000, tdiv at 110000000, tin=divclk, base 0
[    0.310000] Switching to clocksource mct-frc
[    0.362772] FIMD src sclk = 440000000
[    0.362782] s3cfb s3cfb.0: vclk(124968000) should be smaller than 100000000
[    0.362793] s3cfb s3cfb.0: vclk(110000000) should be smaller than 100000000 Hz
[    2.878990] Mali: init_mali_clock mali_clock c07ce5c4
[    3.001016] s3c-sdhci s3c-sdhci.2: clock source 2: sclk_mmc (55000000 Hz)
[    3.028423] s3c-sdhci s3c-sdhci.3: clock source 2: sclk_mmc (55000000 Hz)


Kernel 3.8:

Code: Select all
exynos4_mfc_setup_clock(&s5p_device_mfc.dev, 267 * MHZ);
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby odroid » Wed Aug 07, 2013 10:41 am

You guys found a lot of interesting items..

BTW, is there any possiblity of Dynamic Frequency Scaling on FIMC/MFC/FIMD/HDMI by user land driver in Android platform?
If it is... it might cause a lot of confusion.
User avatar
odroid
Site Admin
 
Posts: 22276
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Wed Aug 07, 2013 11:03 am

Just to add more to the clocks stuff :)

In 3.8 kernel the mct-frc clock is not initialized and used at all. The default sched_clock "step" is 50ms in 3.8 kernel, and 1 ms in 3.0 kernel with mct-frc clock.
You can see the difference in kernel output in 3.0 kernel:

Code: Select all
[    0.200000] print_constraints: LDO26 VDD_MOTOR_3V0: 3000 mV
[    0.200000] print_constraints: BUCK1 vdd_mif: 1100 mV
[    0.200000] print_constraints: BUCK2 vdd_arm: 800 <--> 1500 mV at 1300 mV
[    0.200000] print_constraints: BUCK3 vdd_int: 1125 mV
[    0.200000] print_constraints: BUCK4 vdd_g3d: 850 <--> 1200 mV at 1100 mV
[    0.205000] print_constraints: BUCK5 VDDQ_CKEM1_2: 1200 mV
[    0.205000] print_constraints: BUCK6 1V35: 1350 mV
[    0.205000] print_constraints: BUCK7 2V0: 2000 mV
[    0.305000] print_constraints: BUCK8 3V0: 3300 mV
[    0.310000] Disable Regulator!!
[    0.310000] print_constraints: BUCK9 1V2: 1200 mV
[    0.310000] print_constraints: EN32KHz AP:
[    0.310000] print_constraints: EN32KHz CP:
[    0.310000] s3c-i2c s3c2440-i2c.0: i2c-0: S3C I2C adapter
[    0.310000] s3c-i2c s3c2440-i2c.1: i2c-1: S3C I2C adapter
[    0.310000] s3c-i2c s3c2440-i2c.4: i2c-4: S3C I2C adapter
[    0.310000] s3c-i2c s3c2440-i2c.5: i2c-5: S3C I2C adapter
[    0.310000] Advanced Linux Sound Architecture Driver Version 1.0.24.
[    0.310000] Bluetooth: Core ver 2.16
[    0.310000] NET: Registered protocol family 31
[    0.310000] Bluetooth: HCI device and connection manager initialized
[    0.310000] Bluetooth: HCI socket layer initialized
[    0.310000] Bluetooth: L2CAP socket layer initialized
[    0.310000] Bluetooth: SCO socket layer initialized
[    0.310000] cfg80211: Calling CRDA to update world regulatory domain
[    0.310000] Switching to clocksource mct-frc
[    0.310784] Switched to NOHz mode on CPU #0
[    0.311096] Switched to NOHz mode on CPU #1
[    0.311184] Switched to NOHz mode on CPU #2
[    0.311315] Switched to NOHz mode on CPU #3
[    0.321375] NET: Registered protocol family 2


You have 0.200000, 0.205000, 0.305000, 0.310000 and then immediately after mct is enabled you have 1 ms resolution: 0.310784, 0.311096, ....
With mdrjr's help, I enabled mct on 3.8 kernel, which was then set to what I see now is the xusbxti (24MHz), which was actually much much less than 1ms, and the clock would rollover every 3 minutes. This actually caused some kernel panics when I was debugging mali, whereas there was no panic on 3.0 kernel, where the mct rate was higher (corresponding to 1ms).

Just trying to point out that not all clocks are enabled/set the same as in 3.0 kernel...

EDIT: I got the ms/ns numbers wrong, but you get the idea :)
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby OverSun » Wed Aug 07, 2013 3:08 pm

To make things clear - I'm trying to speed up FIMC conversion now, since it is clearly the slowest thing for current moment. MFC can do 30 fps on 1080p, FIMC is 15 fps on anything >720p already. And it was speeded up when xusbxti clock was raised.

And another thoughts I have looking at the code: generally FIMC is using
.reg_src = { .reg = EXYNOS4_CLKSRC_CAM, .shift = 4, .size = 4 },
.reg_div = { .reg = EXYNOS4_CLKDIV_CAM, .shift = 4, .size = 4 },
so camera clocks, which can be less quick than mfc. It sounds right since fimc in general is a camera input conversion device to something programs can understand.
But for MFC another kind of fimc is used, which is FIMC m2m device, direct memory to memory conversion, and probably it could be hooked to other clocks which can be faster. What do you think, is it possible?
User avatar
OverSun
 
Posts: 1186
Joined: Mon Apr 29, 2013 5:12 pm
languages_spoken: english

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Wed Aug 07, 2013 3:14 pm

I think that refers to registers, not clocks.

M2M is actually a V4L construct, not FIMC.
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Wed Aug 07, 2013 3:23 pm

I've seen in some older files fimc is set to 166MHz...

Code: Select all
+'fimc' device nodes
+-------------------
+
+Required properties:
+
+- compatible : "samsung,s5pv210-fimc" for S5PV210, "samsung,exynos4210-fimc"
+  for Exynos4210 and "samsung,exynos4212-fimc" for Exynos4x12 SoCs;
+- reg : physical base address and size of the device  memory mapped registers;
+- interrupts : FIMC interrupt to the CPU should be described here;
+- samsung,pix-limits : an array of supported image sizes in pixels, for details
+  refer to Table 2-1 in the S5PV210 SoC User Manual.
+
+For every fimc node a numbered alias should be present in the aliases node.
+Aliases are of the form fimc<n>, where <n> is an integer (0...N) specifying
+the IP's instance index.
+
+Optional properties:
+
+- clock-frequency : maximum FIMC local clock (LCLK) frequency;
+- samsung,min-pix-sizes : an array specyfing minimum image size in pixels at
+  the FIMC input and output DMA, in the first and second cell respectively.
+  Default value when this property is not present is <16 16>;
+- samsung,min-pix-alignment : minimum supported image height alignment (first
+  cell) and the horizontal image offset (second cell). The values are in pixels
+  and default to <2 1> when this property is not present;
+- samsung,mainscaler-ext : a boolean property indicating whether the FIMC IP
+  supports extended image size and has CIEXTEN register;
+- samsung,rotators : a bitmask specifying whether this IP has the input and
+  the output rotator. Bits 4 and 0 correspond to input and output rotator
+  respectively. If a rotator is present its corresponding bit should be set.
+  Default value when this property is not specified is 0x11.
+- samsung,cam-if : a bolean property indicating whether the IP block includes
+  the camera input interface.
+- samsung,isp-wb : this property must be present if the IP block has the ISP
+  writeback input.
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby panda84 » Fri Aug 09, 2013 2:52 am

memeka wrote:Mali acceleration
  • /etc/X11/xorg.conf at http://pastebin.com/QCYFYad2
  • 3.0.80
    • glxgears: software renderer, ~120fps with 1 core @ 100%
    • es2gears: EGL_VERSION: 1.4 Linux-r3p2-01rel0, 560-600+fps with ~ 1 core @ 40%, 1 core @ 40-100%
    • glmark2-es2: GL_RENDERER: Mali-400 MP, GL_VERSION: OpenGL ES 2.0, ~ 1 core @ 50-90%, 2 cores @ <20%, score: 140fps Debian, 137fps Ubuntu (fullscreen 80 fps Debian, 73 Ubuntu)
  • 3.8.13
    • glxgears: software renderer, ~ 120fps with 1 core @ 100%
    • es2gears: EGL_VERSION: 1.4 Linux-r3p2-01rel0, 360-380fps with ~ 2 cores @ 30% NOTE: overclocked at 640MHz
    • glmark2-es2: GL_RENDERER: Mali-400 MP, GL_VERSION: OpenGL ES 2.0, ~ 1 core @ 40% (up to 80%, +20% core2 for some tests), score: 90fps (53 fullscreen) NOTE: overclocked at 640MHz


Hi memeka,

at which resolution and under which Window Manager / DE did you achieve that glmark2-es2 scores? I can almost match the 140 points @ 800x600 in openbox, but can even get near at 80 points @ 1920x1080. Maybe you used 1280x720 fullscreen?

Bests,
Diego
panda84
 
Posts: 4
Joined: Tue Aug 06, 2013 2:05 am
languages_spoken: english

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Sun Aug 25, 2013 11:51 pm

They are in HDMI 720p resolution (1280x720).
In Debian + fluxbox.

Updated results on first post with mali fix for 3.8.13 kernel - 91fps in glmark-es2 fullscreen.
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby neagix » Mon Sep 02, 2013 3:36 am

I have published by new benchmarks: http://neagix.blogspot.com/2013/06/the-state-of-hardware-acceleration-on.html

The 3.8 performance bug is fixed for me :)

I have also compiled XBMC on Debian with this script and I am going to give it a shot as well
neagix
 
Posts: 264
Joined: Sun Apr 14, 2013 2:54 am
languages_spoken: english
ODROIDs: U2

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Thu Sep 05, 2013 9:53 am

Updated with XBMC instructions.
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby jeky82 » Sun Sep 08, 2013 11:16 pm

Hi memeka, I am traying to compile xbmc following your instruction on Ubuntu 13, but I get this error..

thank you in advance

Code: Select all
CPP     xbmc/cores/paplayer/SIDCodec.o
CPP     xbmc/cores/paplayer/TimidityCodec.o
CPP     xbmc/cores/paplayer/VGMCodec.o
CPP     xbmc/cores/paplayer/WAVcodec.o
CPP     xbmc/cores/paplayer/YMCodec.o
AR      xbmc/cores/paplayer/paplayer.a
LD      xbmc.bin
/usr/bin/ld: xbmc/windowing/egl/windowing_egl.a(EGLNativeTypeOdroid.o): undefined reference to symbol 'XGetWindowAttributes'
/usr/bin/ld: note: 'XGetWindowAttributes' is defined in DSO /usr/lib/arm-linux-gnueabihf/libX11.so.6 so try adding it to the linker command line
/usr/lib/arm-linux-gnueabihf/libX11.so.6: could not read symbols: Invalid operation
collect2: error: ld returned 1 exit status
make: *** [xbmc.bin] Error 1

jeky82
 
Posts: 35
Joined: Tue Apr 30, 2013 7:58 pm
languages_spoken: italian, english
ODROIDs: odroid u2

Re: Ubuntu U2 projects status DASHBOARD

Unread postby memeka » Sun Sep 08, 2013 11:22 pm

Hi,

Maybe you have the wrong headers, or you don't have the X11 dev package installed...
User avatar
memeka
 
Posts: 3437
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: Ubuntu U2 projects status DASHBOARD

Unread postby jeky82 » Sun Sep 08, 2013 11:30 pm

memeka wrote:Hi,

Maybe you have the wrong headers, or you don't have the X11 dev package installed...


Thank you :) ... I checked for x11 dev, but it is installed, so probabily as you suggested I have the wrong headers, but I do not know how I have to proceed :)
jeky82
 
Posts: 35
Joined: Tue Apr 30, 2013 7:58 pm
languages_spoken: italian, english
ODROIDs: odroid u2

Next

Return to Ubuntu

Who is online

Users browsing this forum: No registered users and 1 guest