This is easier than one would think. Say you have many large shared-memory regions (e.g. queues) with control structures at the start. Bam, all those control

This is easier than one would think. Say you have many large shared-memory regions (e.g. queues) with control structures at the start. Bam, all those control structures, and therefore your atomic accesses, are now sharing cache lines.