// We may only have one actor on a node. If that actor executes something like
// compute -> recv -> compute
// the recv operation will not trigger a "CpuAction::on_state_change". This means
// that the next trigger would be the 2nd compute, hence ignoring the idle time
// during the recv call. By updating at the beginning of a compute, we can
// fix that. (If the cpu is not idle, this is not required.)
// We may only have one actor on a node. If that actor executes something like
// compute -> recv -> compute
// the recv operation will not trigger a "CpuAction::on_state_change". This means
// that the next trigger would be the 2nd compute, hence ignoring the idle time
// during the recv call. By updating at the beginning of a compute, we can
// fix that. (If the cpu is not idle, this is not required.)