Parallel Processing, Parallel Processors

In the past, when processors were more expensive, designers used only one processor in a machine. Today, many computers contain multiple processors. Although it can be a desirable feature, you must be careful when evaluating parallel processing machines. If a computer has four processors (or four cores), it is tempting to say that the machine is four times faster than a computer with only one

processor. Indeed, many computer companies advertise their computers this way.

Can a computer with four processors really do your job four times faster? The answer is that it depends on the job. Consider an example. A computer with two processors has to add two sets of numbers together. Each processor works on one pair of numbers and finishes in half the time of a single processor.

Notice that the second calculation depends on the outcome of the first one. The second one cannot be computed until the first one is finished. Even if one processor is assigned to each calculation, the parallel-processing machine will take just as long as the single processor. On the other hand, if you commonly run multiple

applications at the same time, multiple processors will improve performance because each processor can handle the applications separately.

Massively parallel machines can include thousands of processors. They are used for some highly specialized applications. For example, governments use them to break codes; physicists use them to simulate large-scale events such as nuclear explosions and weather patterns; computer artists use them at special-effects

studios, such as Industrial Light & Magic, to create movies. This concept has been extended with the use of grid computing, where thousands or millions of separate computers are assigned portions of a task.