Optimize driver threading strategy

[Svastits]

On client machines with less CPU cores/available threads, the lifecycle transitions of the robot_manager nodes can fail due to the lack of resources.
This problem should be investigated and optimized if possible

[prasuchit]

@Svastits I just moved to Jazzy and I still needed the RT kernel to connect with the robot since it wasn't working with the generic kernel. Any update on this optimization issue? Please let me know.

[Svastits]

could you share some details about your setup (hardware and software), where it was not working?
We tried it with low-performance hardware and did not find any issues...

Did you enable the configuration of priorities (see wiki) on your machine when this was not working? That is necessary even without the RT patch

[prasuchit]

Hi @Svastits ,

The computer I'm using is a Dell Precision Tower with a 12GB Nvidia RTX A2000 GPU and Nvidia 560.35 driver and a CUDA 12.6 installed. It has 64GB RAM if I remember correctly. Without an RT kernel even the configure command wouldn't work. So I would say it definietly depends on the kernel a lot. Thanks for looking into this. :)