|
unsigned int | flow::async::optimal_worker_thread_count_per_pool (flow::log::Logger *logger_ptr, bool est_hw_core_sharing_helps_algo) |
| Assuming a planned thread pool will be receiving ~symmetrical load, and its UX-affecting (in particular, per-op latency-affecting) operations are largely between processor and RAM: Returns the # of threads to store in that pool for efficient performance. More...
|
|
void | flow::async::optimize_pinning_in_thread_pool (flow::log::Logger *logger_ptr, const std::vector< util::Thread * > &threads_in_pool, bool est_hw_core_sharing_helps_algo, bool est_hw_core_pinning_helps_algo, bool hw_threads_is_grouping_collated) |
| Assuming the same situation as documented for optimal_worker_thread_count_per_pool(), and that indeed the pool now contains that number of running threads: Attempts to optimize thread-core-pinning behavior in that pool for efficient performance. More...
|
|