At the start of every time step, the manager thread sends
a notification to each worker thread to start processing their
work queues and waits for a response. The workers then send
notifications back to the manager to indicate completion of
the cycle, and wait for another notification to start. Waiting
in this context implies a spin-wait that prevents releasing a
CPU when no work remains. Since each worker thread and
the manager thread has exclusive access to a single CPU,
this allows instantaneous start up once the new cycle begins.
The RedHawk tools discussed earlier ensure that absolutely
nothing else runs on these CPUs, including operating system
interrupt handlers and the system timer.