Cuda threads invocation from linux kernel space threads

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,

I am using Nvidia’s Quadro P400.
I am running Cuda 10.0 on Ubuntu 18.04 with driver version 410.
The development IDE is Nsight on Eclipse.
I am new to the GPU environment but so far everything is running smooth thanks to the tutorials.

I did notice that Cuda drives the developer to code in the linux user space.
Basically, the host threads invoked (in my .cu file) are all in the linux user space.
Consequently, the Cuda threads are invoked from the linux user space threads.

So my questions are -

  1. Can I use nvcc to compile linux kernel space code (specifically the linux driver code) ?
  2. More importantly, can the Cuda threads be invoked from the linux kernel space threads ?

Regards,
Akhil Jain