search

ms-van3t-devs / ms-van3t

A multi-stack, ETSI compliant, V2X framework for ns-3.
GNU General Public License v2.0
103 stars 34 forks source link

Slow running performence #28

Open melihyazgan opened 6 months ago

melihyazgan commented 6 months ago

Hi Carlos, I am coming up with another Issue, but this time no bug=) I have really quite slow performance while the simulation is running. Do you have any Idea or can you give hints? Below is my setup:

Ubuntu 20.04.6
Intel® Core™ i7-7800X CPU @ 3.50GHz × 12
GeForce GTX 1080 Ti x 2

Greetings Melih

carlosrisma commented 6 months ago

Well, CARLA itself is a quite demanding software on its own so there's not much things to do from our side to speed-up simulations. If you are using the NR-V2X module that would be the worst combination, being NR-V2X quite slow as well (even with SUMO). The only thing to point out is to make sure that CARLA is indeed running on the dedicated GPU with nvidia-smi. The other thing to consider is that if not entirely necessary you can setup OpenCDA to not show the camera and LiDAR views from the configuration YAML inside path/to/OpenCDA/opencda/scenario_testing/config_yaml (refer to this link for more information). Lastly, the simulation speed will be affected significantly if you are using ML-based perception as it takes a few tens of milliseconds for each image (4 rgb_cameras per CAV * number of CAVs) on each simulation step.

Hope this gives you some hints :)