Great, we've been using performance counters to characterize what's been going on inside caches at runtime and I'm really glad Intel is taking time to add real

Great, we've been using performance counters to characterize what's been going on inside caches at runtime and I'm really glad Intel is taking time to add real monitoring features to cache contention and use. These factors are increasingly becoming a critical part of system performance and our operating system and hypervisor schedulers are woefully underprepared to deal with it. That all said, I'm sure the OS level interfaces are likely to take some time to become available and even more to become usable, the perf_events_open syscall on Linux is a clusterfuck and performance counters have been around for years. This stuff may only be usable from kernelspace for awhile.