A query about the computational performance on PX2 AutoChauffeur Platform

Autonomous Vehicles DRIVE PX 2 General
1

Hi all,

I experienced a problem about the computational performance on PX2 AutoChauffeur. I deployed the latest opencv libraries and ran a program which calculates the dense optical flow using CalcOpticalFlowFarneback() method. I expect the frame rate should be at least 10x better than running
on my laptop, but the truth is that the frame rate is even worse. Should I install any driver for improvement or is there any other solution I should look into?

THANK YOU!

2

Dear infohyu7s,
Does your application uses GPU for computation? If so, what is the GPU you are using on laptop?

3

I wrote the ordinary cpp program and compiled & executed directly on the platfrom. I think I didn’t use GPU if I need to do something with my code. Now I know the next step of my work, thank you for your reminder.

4

Dear infohyu7s,
Single ARM core is not powerful compared to your laptop CPU core. You will get good performance if you make use of GPUs

5

When I reinstall OpenCV4.0-pre with cuda8.0 using cmake there is an error:

[ 9%] Built target gen_opencv_python_source
Makefile:160: recipe for target ‘all’ failed
make: *** [all] Error 2

I searched online for hours but there is no proper solution. How should I deal with it? Thank you.

6

I uninstall the ccache and now it moves on.