The Reorder Core from Rambus reorders requests based on first on priority and second on throughput optimization.
Throughput optimization includes moving same bank/same row requests next to each other, same bank/ different row requests away from each other, moving reads next to reads and write next to writes.
The core can be used to optionally enforce data coherency by preventing any same row requests from passing over each other.
The core can also optionally disable intra-port (within a port) reordering. Disabling intra-port reordering ensures that the requests on each port are always executed in the same order. In this case only inter-port (between ports) reordering is allowed. The core can be used in a single port mode or a multi-port mode in conjunction with Rambus Multi-Port Front- End Core.
To minimize latency, the core’s Reorder Queue is bypassed when empty. This enables requests to flow directly through the core when no outstanding requests have been queued up.
Rambus also provides IP Core customization services. Contact Rambus for a quote.