I wonder if hardware support for memory duplication is feasible. A de-duplication controller could use DMA and only lock the bus when other devices (eg. the CPU

I wonder if hardware support for memory duplication is feasible. A de-duplication controller could use DMA and only lock the bus when other devices (eg. the CPU) isn't using it. It could then scan for duplicates with virtually no overhead and interrupt the OS when it detects a hit. As a further enhancement, could hardware pick up on memory writes and detect duplication on the fly? What if a device kept track of a hash of all pages? The concurrency possibilities in hardware allow for O(1) operation (if there is enough silicon area available). Partial out-of-order page writes would have to be queued for hash re-computation, but successful de-duplication is less likely in these cases anyway.