I wonder if it is possible to call get mapped page-locked memory using cudaGetDevicePointer once.
The problem I have a collision detection application written in cuda and large amount of data need to be sent to device. Most of the data are actually the same from frame i to frame i+1. The only data that changes is the world matrices of each object in the scene. These matrices need to be copied to GPU to correctly transform every object and perform collision detection.
So can I call cudaGetDevicePointer once for some pointers?