No. cudaHostRegister just pins the CPU memory of an already allocated CPU buffer. It is not dependent on the current GPU. Host memory is always accessible by all GPUs in cudaMemcpy calls.
cudaHostRegister
cudaMemcpy