extreme low FPS after switching to the X.org driver

[ican9595]

Switching to the X.org driver did the trick. The window now shows up, but we have like 0.8 FPS in the simulation... :'-(

Originally posted by @BenedictBauerHHN in #37 (reply in thread)

Thank you Thijs for the support. As requested a new issue regarding to the low FPS Problem in the simulation.

When we switch from the NVIDIA Driver 460 to the X.Org driver to show the simulation with the contest environment in docker our fps rate drops significantly from approximately 40 fps (on the machine (without docker) with the NVIDIA Driver) to 0.8 fps (in the docker container with x.org driver). We also tested the simulation with the x.org driver without docker and had about 1 fps.

How can we get a better performance while using in the contest environment with docker?

[ThijsRuigrok]

In the branch nvidia-docker https://github.com/FieldRobotEvent/Virtual_Field_Robot_Event/compare/nvidia-docker we used docker with the nvidia runtime. Probably it works slightly different on your setup, but I hope that our code will give you an idea on how to proceed.

[ThijsRuigrok]

If you have a GPU on your computer you can try to disable the software rendering for the a container. Long story short, on systems without a GPU the lidar data was not always rendered correctly (was not appearing). When forcing the system to use software rendering by setting LIBGL_ALWAYS_SOFTWARE=1 the lidar was rendered correctly, but but rendering became very inefficient.

When you have a GPU with the x.org driver you might not need the software rendering(I have not verified this). You can try do remove the line in the docker file https://github.com/FieldRobotEvent/Virtual_Field_Robot_Event/blob/master/competition_environment/a_container/Dockerfile that says ENV LIBGL_ALWAYS_SOFTWARE=1, and then build the a_container(simulation container) manually (docker build -t fieldrobotevent/a_container_2021 a_container. Afterwards you can start the competition environment(start_competition_environment.sh), enter one of the containers (docker exec -it fre_a_container_1), In the container, source ros_entrypoint.sh and then rostopic echo front/scan. If some of the values under the tag ranges are not 0 or inf, the lidar is rendered correctly.