How Parallel Processing Works

To understand parallel processing, we need to look at the four basic programming models. Computer scientists define these models based on two factors: the number ofinstruction streamsand the number ofdata streamsthe computer handles. Instruction streams arealgorithms. An algorithm is just a series of steps designed to solve a particular problem. Data streams are information pulled from computer memory used as input values to the algorithms. The processor plugs the values from the data stream into the algorithms from the instruction stream. Then, it initiates the operation to obtain a result.

Single Instruction, Single Data(姐姐D)computershave one processor that handles one algorithm using one source of data at a time. The computer tackles and processes each task in order, and so sometimes people use the word "sequential" to describe SISD computers. They aren't capable of performing parallel processing on their own.

Multiple Instruction, Single Data(MISD)计算机有多个处理器。每一个过程or uses a different algorithm but uses the same shared input data. MISD computers can analyze the same set of data using several different operations at the same time. The number of operations depends upon the number of processors. There aren't many actual examples of MISD computers, partly because the problems an MISD computer can calculate are uncommon and specialized.

Single Instruction, Multiple Data(SIMD) computers have several processors that follow the same set of instructions, but each processor inputs different data into those instructions. SIMD computers run different data through the same algorithm. This can be useful for analyzing large chunks of data based on the same criteria. Many complex computational problems don't fit this model.

Multiple Instruction, Multiple Data(MIMD) computers have multiple processors, each capable of accepting its own instruction stream independently from the others. Each processor also pulls data from a separate data stream. An MIMD computer can execute several different processes at once. MIMD computers are more flexible than SIMD or MISD computers, but it's more difficult to create the complex algorithms that make these computers work.Single Program, Multiple Data(SPMD) systems are a subset of MIMDs. An SPMD computer is structured like an MIMD, but it runs the same set of instructions across all processors.

Out of these four, SIMD and MIMD computers are the most common models in parallel processing systems. While SISD computers aren't able to perform parallel processing on their own, it's possible to network several of them together into acluster. Each computer's CPU can act as a processor in a larger parallel system. Together, the computers act like a single supercomputer. This technique has its own name:grid computing. Like MIMD computers, agrid computingsystem can be very flexible with the right software.

How does a parallel processing system solve computational problems? Find out in the next section.

Individually, each processor works the same as any other微处理器. The processors act on instructions written in汇编语言. Based on these instructions, the processors perform mathematical operations on data pulled fromcomputer memory. The processors can also move data to a different memory location.

In a sequential system, it's not a problem if data values change as a result of a processor operation. The processor can incorporate the new value into future processes and carry on. In a parallel system, changes in values can be problematic. If multiple processors are working from the same data but the data's values change over time, the conflicting values can cause the system to falter or crash. To prevent this, many parallel processing systems use some form ofmessagingbetween processors.

Processors rely on software to send and receive messages. The software allows a processor to communicate information to other processors. By exchanging messages, processors can adjust data values and stay in sync with one another. This is important because once all processors finish their tasks, the CPU must reassemble all the individual solutions into an overall solution for the original computational problem. Think of it like a puzzle -- if all the processors remain in sync, the pieces of the puzzle fit together seamlessly. If the processors aren't in sync, pieces of the puzzle might not fit together at all.

There are two major factors that can impact system performance:latencyandbandwidth. Latency refers to the amount of time it takes for a processor to transmit results back to the system. It's not good if it takes the processor less time to run an algorithm than it does to transmit the resulting information back to the overall system. In such cases, a sequentialcomputersystem would be more appropriate. Bandwidth refers to how much data the processor can transmit in a specific amount of time. A good parallel processing system will have both low latency and high bandwidth.

有时,并行处理不超过equential computing. If it takes too long for the computer's CPU to reassemble all the individual parallel processor solutions, a sequential computer might be the better choice. As computer scientists refine parallel processing techniques and programmers write effective software, this might become less of an issue.

To learn more about parallel processing, follow the links on the next page.