Abstract
Persistent memory (PM) obviously can make many cloud and enterprise applications run much faster, since it offers DRAM-like speeds in an inexpensive, non-volatile form. The problem is how to attach it in a way that takes full advantage of the added performance benefits while allowing for a variety of use cases. An FPGA-based solution is both fast and flexible, offering a competitive advantage over other solutions. It can provide a coherent interface, perform both acceleration and computational functions, keep up with the latest PM implementations, and handle a variety of memory hierarchies including on-chip memories, external memories, and system in package memories. The result is a cost-effective memory accelerator for demanding cloud and enterprise applications.