How does the barrier know when all the threads reach the given point? If messages are passed between the threads, do we have to account for this communication in the overhead of parallelizing?
bpr
@llcoolj, We'll visit barrier design next month in when discussing how synchronization is implemented. So have your questions ready.
In brief, there is overhead in any synchronization, be it lock, barrier, et cetera, that limits the scalability of the system. However, there are efficient designs to mitigate these costs.
How does the barrier know when all the threads reach the given point? If messages are passed between the threads, do we have to account for this communication in the overhead of parallelizing?
@llcoolj, We'll visit barrier design next month in when discussing how synchronization is implemented. So have your questions ready.
In brief, there is overhead in any synchronization, be it lock, barrier, et cetera, that limits the scalability of the system. However, there are efficient designs to mitigate these costs.