2020-09-26 06:54 PM
I have xfce running very well in the V1.2 Yocto Distribution system on the stm32mp157c disco board. The xfce desktop is very responsive and uses about 100M of memory. The xfce desktop using the latest V2.0.0 Distribution image however is very sluggish and even a simple mouse movement will cause near 100% cpu utilization. Memory usage is around 200M or more with the Xorg server alone using up 16M of memory. It would appear that the new image is not accelerated. I see similar behavior with the x11 image (matchbox desktop) so the issue is not the desktop but with the underlying system. Is there something setup parameter I am missing? I also noted that the weston image seems to run fine!
2020-10-08 07:28 AM
Hello @STild.1 ,
Thank you for your message.
I reproduced the issue and I escalated the point to the development team.
I keep you posted.
Regards,
Armand
2020-10-23 08:52 AM
Hi @STild.1 ,
According to the x11 server trace, it seems the server is not compliant with the way of working of drivers for mapping/unmapping buffer with kernel 5.4.
But actually the x11 image is not really suitable for embedded systems, it is only for examples. If you don't really need to use this image I recommend you to use the wayland weston image which is officially supported by the STM32MP1.
Regards,
Armand
2020-10-23 06:03 PM
ArmandG,
I appreciate you looking into the matter of X11 support. Actually we do need X11 for some of the post-processing/result display that we want to do on the LCD screen. We have these apps working just fine on the 4.19 kernel system and was hoping for an upgrade, especially with later (post python 3.5) support. I will work on trying to port to Wayland/Weston but it will take a LOT of work since the current python libraries we are working with don't support this framework. I politely disagree with your assessment of X11 on your hardware, X11 has huge advantages for doing for example X-redirection of apps to a desktop for troubleshooting, etc. X support works very well on your hardware with the 4.19 kernel system. This sort of app network redirection cannot be done on Wayland/Weston and no plans seem to be in the works for re-direction support in this framework. I have done X on even less constrained embedded systems in the past - many science apps we work with require it. I am also working on the later STM32 kernels (5.7 and 5.8) with the Digikey framework (Robert Nelson) but don't have X working yet due to several configuration issues. I'll give an update when/if we are successful with this.
Scott Tilden, PhD
U of Arizona