von Neumann structure
Traditional computers use the von Neumann (Von Neumann) structure, also known as the Princeton structure, as a memory structure that incorporates program instruction memory and data memory together. Von Neumann structure of the computer its program and data common one storage space, program instruction store address and data storage address point to the same memory different physical location, with a single address and data bus, program instruction and data width is the same. When the processor executes the instruction, it takes the instruction decoding from the storage and then the operand to perform the operation, even if the single instruction consumes several or even dozens of cycles, the bottleneck effect will appear on the transmission channel when the high-speed operation is taken.
As shown in 1-3, the von Neumann structure of the computer is composed of CPU and memory, the program calculator (PC) is the CPU internal instruction and data storage location register. The CPU passes the address information provided by the program counter, addresses the memory, finds the required instruction or data, decodes the instruction, and finally executes the operation specified in the instruction.
In this architecture, the program counter is only responsible for providing the instructions and data required for the execution of the program, without determining the program flow. To control the process, you must modify the directives.
Harvard structure
The Harvard (HARVARD) structure is a memory structure that separates program instruction storage from data storage. Harvard structure is a parallel architecture, its main feature is to store programs and data in different storage space, that is, program memory and data memory is two independent memory, each memory independent address, independent access. Corresponding to two memory is the system of 4 sets of bus: The program's data bus and address bus, data bus and address bus. This separate program bus and data bus allows for simultaneous acquisition of the instruction word (from program memory) and operands (from the data memory) within a machine cycle, thus increasing execution speed and increasing the throughput rate of the data by 1 time times. Because the program and the data memory are in two separate physical space, the reference and the execution can completely overlap.
1-4, the Harvard structure of the computer is composed of CPU, program memory and data memory, program memory and data memory using different bus, thus providing a large memory bandwidth, so that the movement and exchange of data more convenient, especially to provide high digital signal processing performance.
A Harvard-structured CPU typically has high execution efficiency. There are many CPUs and microprocessors using the Harvard structure, in addition to all DSP processors, Motorola's MC68 series, Zilog's Z8 series, Atmel's AVR series and Arm's ARM9, ARM10 and AR
M11 and so on.
Von Neumann structure and Harvard structure