This strategy addresses the recurring theme of identifying dependencies when designing parallel programs. Since we can only parallelize tasks if they are independent of each other, we can schedule our tasks more efficiently if we can precompute a dependency graph of some kind between tasks. So, if task A depends on task B and task B depends on tasks C and D which are independent, our scheduler will know it cannot schedule task A before completing task B and cannot schedule task B before tasks C and D. However, because C and D have no dependencies, they can both be scheduled simultaneously.
This strategy addresses the recurring theme of identifying dependencies when designing parallel programs. Since we can only parallelize tasks if they are independent of each other, we can schedule our tasks more efficiently if we can precompute a dependency graph of some kind between tasks. So, if task A depends on task B and task B depends on tasks C and D which are independent, our scheduler will know it cannot schedule task A before completing task B and cannot schedule task B before tasks C and D. However, because C and D have no dependencies, they can both be scheduled simultaneously.