What can be parallelized in a single processor system?

In a single-processor system, "processor and device", "processor and channel", and "device and device" can be parallelized. The formation of modern operating systems is the result of channels. Channels are devices that can control input and output independently of the CPU. When the processor is performing operations, the channel can perform I/O operations at the same time. Therefore, the processor and the channel are parallel; When the processor is operating, the peripherals can perform input/output operations at the same time, so the processor and the device can be parallelized; and the devices can obviously be parallelized.

The operating environment of this tutorial: Windows 7 system, Dell G3 computer.

A computer system that includes only one computing processor is called a single-processor system.

In a single-processor computer system, there are at most 1 running state and at least 0; there are at most N waiting states and at least N-1; and at most N-1 and at least 0 ready states.

The single-processor system has only one instruction pipeline and one multi-functional operating component. Each clock cycle "fetches instructions" and "analyzes" completes one instruction. In many pipeline processors, the number of pipeline stages in the instruction pipeline is k=4; it mainly decomposes the execution process of an instruction into four stages: "fetching instructions", "analyzing", "executing" and "writing results". The function to be executed by the instruction is mainly completed in the multi-functional operating component and is completed in the "execution" section. The multi-functional operating components of most pipeline processors adopt a pipeline structure. Some simple instructions can be completed in the "execution" pipeline segment in just one clock cycle, while more complex instructions often require multiple clock cycles. In addition, there are also the effects of conditional transfer, etc.; therefore, the average number of instructions executed by a general pipeline scalar processor per clock cycle is less than 1, that is, its instruction level parallelism ILP<1.

In a single-processor system, what can be parallelized are: processor and device, processor and channel, device and device; but processes cannot be parallelized,

• In a single-processor system, in order to improve system efficiency, multiple processes in different states are maintained in the memory (this is called concurrent execution of processes), but they occupy the processor at the same time. But there is only one process, so the processes are not parallel;

• The formation of modern operating systems is the result of channels. Channels are devices that can control input and output independently of the CPU. When the processor is performing operations, the channels can perform I/O operations at the same time, so the two are also parallel;

• When the processor is performing operations, the peripherals can perform input/output at the same time Operations, therefore, also produce parallelism;

• Obviously, parallelism between devices can also be achieved.

Extended knowledge: The difference between a single-processor system and a multi-processor system

Early computer systems were sequential processing machines based on a single processor . Programmers write code for serial execution and let it be executed serially on the CPU. Even the execution of each instruction is serial (fetching instructions, fetching operands, performing operations, and storing results). In order to improve the speed of computer processing, the Lenovo memory system and the pipeline system were first developed. The former proposed the idea of ​​​​data-driven, and the latter solved the problem of serial execution of instructions. Both of them were the initial development of computer parallelism. example. With the advancement of hardware technology, parallel processing technology has developed rapidly. Computer systems are no longer limited to single processors and single data streams, and various parallel structures have been applied.

Currently computer systems can be divided into the following four categories.

(1) Single instruction stream single data stream (SISD). A processor executes a single stream of instructions on data in memory.

(2) Single instruction stream multiple data stream (SIMD). A single instruction flow controls the simultaneous execution of multiple processing units. Each processing unit includes a processor and related data storage. One instruction actually controls different processors to operate on different data. Vector machines and array machines are representatives of this type of computer system.

(3) Multiple instruction stream single data stream (MISD). A data stream is transmitted to a group of processors and the processing results are finally obtained through different instruction operations of this group of processors. Research on this type of computer system is still in the laboratory stage.

(4) Multiple instruction stream multiple data stream (MIMD). Multiple processors execute different instruction streams simultaneously on different data sets. MIMD systems can be divided into two categories: tightly coupled MIMD systems with shared memory and loosely coupled MIMD systems with distributed memory.

For more related knowledge, please visit the FAQ column!

The above is the detailed content of What can be parallelized in a single processor system?. For more information, please follow other related articles on the PHP Chinese website!