How can I use Pined memory for using cuda in jetson xavier

Hello Everyone.
I am using jetson xavier for doing person detection and person tracking and person re_id and gesture recognition with ros based ORB_SLAM and Lidar SLAM.
But I heard that the xavier using shared memory for cpu and gpu operation.
So I think that .If I can pin the momory for cuda then Xavier can perform better than now.
So I just wonder how to pin the momory for using gpu(cuda).
Is there any simple way to pin the memory of Xavier?
Is there only way that doing cuda coding?

Please answer me…