Currently, our implementation of std_thread is to assign a fixed number of rows to each thread. It will simply cause a workload unbalanced issue. For better performance, we can use the hread_pool strategy to dynamically assign the row (work) to the idle worker in thread_pool.
Currently, our implementation of std_thread is to assign a fixed number of rows to each thread. It will simply cause a workload unbalanced issue. For better performance, we can use the hread_pool strategy to dynamically assign the row (work) to the idle worker in thread_pool.