Hello Shuo Wang,
Maybe I should consider the time cost of using DoPE on Genode, because my work should be finished before February, 2015.
I have just dusted off the Genode version of DOpE and cleaned it up a bit. If you like to give it a try, you can find it here:
https://github.com/nfeske/dope
- Which version of Genode are you using?
We're using version 13.02.
- Are you using Qt4 or Qt5? (If you are still using Qt4, I would
recommend you to switch to Qt5)
Qt 4.7.4 was using. I tried to replace Qt4 with Qt5, but failed. The reason maybe the ways I used are wrong because of my lack of knowledge about that. So, could you give me some tips or tutorial about Qt4's updating to Qt5? Sorry to bother you, Norman.
Qt5 is supported since version 13.08. To try it out, you might consider updating to a newer version of Genode. In Genode 14.02, you find plenty of examples at 'libports/run/qt5_*.run'. Using Qt5 is actually not very different from using Qt4. You can find more information about switching to Qt5 at the following link:
http://www.genode.org/documentation/release-notes/13.08#Qt5_available_on_all...
- Which Genode base platform are you using?
Fiasco.OC, because our prototype and research are all based on microkernel architecture till now.
Are you using Fiasco.OC on the ARM, x86_32, or x86_64 architecture? If ARM, which SoC is used?
- Have you tried to run your GUI application on base-linux? If yes, have you compared its performance to the same application compiled for
native Linux w/o Genode?
I would like to carry some experiments before giving an answer to you.
- On base-linux, you could use normal Linux profiling tools such as
oprofile to find out where the CPU time goes.
It's good way, but I should consider how to make sure that Fiasco.OC-based Genode's performance is improved.
Trying out the same scenario on different kernels is a good way to see whether the performance problem is related to the specific kernel or somewhere in the generic code (which is by large the majority of the code). If you can reproduce the performance problem on base-linux, you can debug it there (using Linux' profiling and tracing tools). Once you solved the problem on Linux, it will most likely be fixed on Fiasco.OC as well. On the other hand, if you observe the performance on base-linux to be fine, the problem lies somewhere in the Fiasco.OC-specific code.
By testing your program on additional kernels (e.g., base-hw, base-nova), different kernel configurations, or CPU architectures, you might be able to further cross-correlate between the varying factors of those experiments. With Genode, it is quite easy to conduct such experiments, which might give you additional hints to encircle the problem.
I'm so sorry for being not able to do this, because parts of source code of our prototype could not be public. I really appreciate you help, and if some day our team leader agree to open source code to the public, I would like to invite you to our team.
:-)
How about investigating the performance problem using the examples in Genode's source tree then? Do you experience the same poor performance for those? For example, how does 'make run/textedit' perform on your platform?
Best regards Norman