[macros]

There are people exploring pushing more computation down to the device layer to avoid shuffling the bits around: https://www.usenix.org/conference/fast23/presentation/yang-z...

DPUs are also probably a piece you are missing. Vendors are starting to position these as almost PCIe arbiters with access to both nvme devices and GPUs without the need to involve the main cpu to move data between them.

[_8j50]

Cool stuff, I mean at least data comparison/matching would be great but it's pretty wild to think what it would be like if the filesystem itself was implemented by storage firmware as a logical abstraction.