Speaking of drivers on the GPU: https://github.com/enfiskutensykkel/ssd-gpu-dma

> Build userspace NVMe drivers and storage applications with CUDA support

> libnvm can be linked with CUDA programs, enabling high-performance storage access directly from your CUDA kernels. This is achieved by placing IO queues and data buffers directly in GPU memory, eliminating the need to involve the CPU in the IO path entirely.