How to copy memory from Kernel space straight to the GPU? (using gdrcopy)

Accelerated Computing GPU-Accelerated Libraries
1

I want to enable gdrcopy functionality in kernel space and copy memory from Kernel space straight to the GPU. As an amateur coder, I am wondering if this is even possible.

I realize I will need to Write a Kernel module, and allocate memory in the Kernel and the GPU and implement functions from gdrcopy. But am unsure how to tackle this process, any help would be great!