> I think it might be possible to create a build system that can automatically
> collect the profiling data (branch stall cycles and data stall cycles)
> and use the branch likely and prefetch instructions to auto-optimized the
> critical path code.
Recent versions of Clang and GCC actually support profile guided optimizations using gcov format files to trace executions. There are a number optimizations that the compiler can determine profitable knowing how something typically executes. They don't use sampled or simulated metrics though, and I'm not sure if they add prefetch or non-temporal optimizations.
The ones you mentioned aren't exactly equivalent to cache partitioning (etc...) though. Partitioning allows explicit allocation and prioritization of a contended resource instead of only improving utilization for a single process. So, for example if two threads/processes have an 8MiB working set, and they run on the same cpu, they can easily step on eachother's data in the cache. If you partition it though, less frequently used data doesn't have to step on more frequently used data.
The ones you mentioned aren't exactly equivalent to cache partitioning (etc...) though. Partitioning allows explicit allocation and prioritization of a contended resource instead of only improving utilization for a single process. So, for example if two threads/processes have an 8MiB working set, and they run on the same cpu, they can easily step on eachother's data in the cache. If you partition it though, less frequently used data doesn't have to step on more frequently used data.