
58 High Performance Programming for Soft Computing
passing”. One of the greatest advantage of this type of memory is that it is
scalable to the number of computers on the network.
The hybrid distributed memory shown in Fig. 2.24 consists of multiple
processors UMA, NUMA and GPU (Graphics Processing Unit) connected by
a data network and is known for mixing different processing technologies,
including graphics coprocessors (Patterson 2012).
Fig. 2.23 Distributed memory.
Fig. 2.24 Hybrid distributed memory.
2.7 Measuring Performance
Modern computers are very complex because there is a wide range of
performance improving techniques at the hardware level, ...