Not every problem is worth trying to parallelize on many cores. If the problem is too small, the excessive overhead in communication makes it so that the performance gains are effectively moot and in some cases, can actually hurt performance.
This comment was marked helpful 0 times.
martin
This doesn't truly reflect the usage of the 32-processor machine when the problem size is small enough. The ratio between by time consumed by communication and computation can vary a lot due to the size of the problem as well as the number of processors in the machine.
Not every problem is worth trying to parallelize on many cores. If the problem is too small, the excessive overhead in communication makes it so that the performance gains are effectively moot and in some cases, can actually hurt performance.
This comment was marked helpful 0 times.
This doesn't truly reflect the usage of the 32-processor machine when the problem size is small enough. The ratio between by time consumed by communication and computation can vary a lot due to the size of the problem as well as the number of processors in the machine.
This comment was marked helpful 0 times.