The processing capacity of the processor has a variety of indicators and parameters. It is usually measured by the fastest million instructions per second (MIPS). For processors with vector processing capabilities, the maximum number of million floating point processing results (MFLOPS) that can be given per second is used to measure. In addition, the processing data rate (PDR) is often used to evaluate the processing capabilities of the processor. The processing data rate (PDR) is defined as the product of the average number of bits transmitted per instruction and the average instruction processing rate. In the early computers, the structure of the processor was centered on the arithmetic unit, and most of them used serial work. Data input-output (I/O) transmission must go through the arithmetic unit. As solid-state electronic devices replace vacuum electronic devices, the data processing capability of the central processing unit (CPU) has been greatly improved. The data transfer speed of electromechanical peripherals can no longer be compared with it. Therefore, the transmission control component of the input-output operation is separated from the central processing unit into an interface component, a channel or a direct memory access (DMA) component, and the design technology of the interrupt system appears, thereby effectively enhancing the processing capacity of the processor. At the same time, the processor structure has changed to center on the main memory. After that, the bus structure that interconnected the various components of the processor appeared. With the progress of microelectronics technology and the development of computer system structure, large-scale integrated circuits can be used to form processors of different structures and suitable for different purposes, such as array processors, vector processors, array processors, database processors, Input-output processors and microprocessors that build the entire processor on several silicon chips, etc.
The operation of the processor is to first input the user program and data to the main memory (main memory) or auxiliary memory through the input-output device. The central processing unit accesses instructions from the host, completes the interpretation of the instructions, and performs control operations; if it is an arithmetic instruction, it also needs to access the data from the host, and the arithmetic unit completes the calculation. The result is usually temporarily stored in the arithmetic unit or sent back to the main memory.
Executing the program
The process of the processor executing the program involves input-output operations, main memory-auxiliary memory information exchange, and these all go through the input and output interface components. There are three ways to exchange this information between the processor and the outside world. ①Interrupt mode: program I/O. Every transmission of a bit group (such as a word or byte) generates an interrupt, and the CPU executes the corresponding interrupt program to complete. This method is mainly used for slow input-output devices. ②Direct memory access (DMA) mode: Under the control of the hardware circuit, directly complete the exchange of the amount of information specified by an input-output instruction between the fast input-output device and the main memory. ③Channel control mode: each channel has its own channel program to realize the information exchange between the main memory and the input-output device specified by the input-output instruction.
From the perspective of system structure, according to the relationship between the instruction flow executed by the processor and the data flow related to the instruction flow, there is Single Instruction Stream Single Data Stream (SISD) Processor, single instruction stream multiple data stream (SIMD) processor and multiple instruction stream multiple data stream (MIMD) processor. The program of the SISD processor is executed according to a single instruction sequence, and the operating data is also processed one by one according to a single sequence determined by the corresponding instruction. Most processors fall into this category. SIMD and MIMD processors are also called parallel processors. The purpose of the parallel processor is to improve the data processing capability of the processor. The SIMD processor mainly deals with vector data, so it is also called a vector processor. Among them, a processor composed of a single instruction execution component and multiple identical arithmetic processors is called an array (type) processor, such as the ILLIAC-IV in the United States. A SIMD processor that organizes instruction components (called advance control) and arithmetic functional components in a production assembly line is called an assembly line processor, such as the processor of the "Galaxy" computer that was successfully developed in 1983 in China. The Lenovo processor is a SIMD processor with Lenovo memory retrieved by content as the main feature. As for the MIMD processor, it is actually a multi-processor system. It is a system in which multiple identical processors are coupled with each other through a common main memory to form a multi-processing capability.
Processors can be classified according to their functions in the computer system. In general, the instruction system of the processor can reflect the strength of the processor's function and its scope of application. The general-purpose central processing unit has a strong command function and is suitable for various fields or one or several fields of scientific computing, data processing, business applications, and transaction management. The instruction system of some processors has only partial functions, and is often named after its purpose. ① Input-output processor: interprets and executes input-output instructions, has a certain character processing capability, it completes input-output operations and equipment control operations. ②Communication control processor: realize the communication between each processor in the computer network and coordinate their operations. ③Support and maintenance processor: with system console function, it can realize system maintenance and fault diagnosis. ④ Array processor: It is structurally suitable for array and matrix operations, especially signal processing algorithm operations, and can greatly enhance the vector processing capability of the system when it is connected with a pre-processor or host. In addition, there are: a database processor with database management functions; a processor that implements virtual memory paging, etc.