The growth in computing power made available by modern computers has resulted in us facing computational problems of increasing complexity in relatively short time frames. Until the early 2000s, complexity was dealt with by increasing the number of transistors as well as the clock frequency of single-processor systems, which reached peaks of 3.5-4 GHz. However, the increase in the number of transistors causes the exponential increase of the power dissipated by the processors themselves. In essence, there is, therefore, a physical limitation that prevents further improvement in the performance of single-processor systems.
For this reason, in recent years, microprocessor manufacturers have focused their attention on multi-core systems. These are based on a core of several physical processors that share the same memory, thus bypassing the problem of dissipated power described earlier. In recent years, quad-core and octa-core systems have also become standard on normal desktop and laptop configurations.
On the other hand, such a significant change in hardware has also resulted in an evolution of software structure, which has always been designed to be executed sequentially on a single processor. To take advantage of the greater computational resources made available by increasing the number of processors, the existing software must be redesigned in a form appropriate to the parallel structure of the CPU, so as to obtain greater efficiency through the simultaneous execution of the single units of several parts of the same program.