Create host matrix with pinned memory

asked 2015-09-01 20:33:31 -0600

pervk gravatar image

When using a GPU for processing, NVIDIA (and others) recommend that any host memory used to transfer data between the CPU should be allocated as 'pinned' not pageable. This can be in the CUDA world by using, for example cudaMallocHost. What would be the best method do something like that with opencv APIs ?

edit retag flag offensive close merge delete
add a comment