Vector processor-Overview

Source: Internet
Author: User

Adopts parallel processing structures such as advanced control and overlapping operation technology, computing pipeline, and cross-access parallel memory.

It plays an important role in improving the computing speed. However, the parallel processing potential cannot be fully utilized during actual operation. Vector operations are suitable for the structural characteristics of assembly line computers. Combining vector parallel computing with pipeline structure
To a large extent, we can overcome the disadvantages such as large command processing capacity, uneven storage access, serious waiting, and poor flow in pipeline computers, and give full play to the potential of parallel processing structures, significantly improves the computing speed.

Volume computing is a simple type of parallel computing. It is widely used and easy to implement and use. Therefore, vector processing machines (SVM) have developed rapidly. Ti
ASC (1972) and CDC
Star-100 (1973) was the first vector giant computer in the world ). By the end of 1982, there were about 60 giant machines in the world, most of which were vector machines. Yu, China
757 machines and "Galaxy" machines that were successfully developed in per second are also vector machines.

Vector Machines are suitable for Linear Planning, Fourier transformation, filtering calculation, matrix,
Linear Algebra, partial differential equations, integral and other mathematical problems, mainly solving meteorological research and weather forecasting, aerospace and aerospace aircraft design, atomic energy and nuclear reaction research, Geophysical Research, seismic analysis, large-scale engineering
Design, as well as large-scale computing problems in the fields of large-scale simulation of social and economic phenomena.

Vector operations
In normal computers, the basic operation object of machine commands is scalar, while vector machines have complete functions in addition to scalar processing.

Each component of a vector
A new vector can be generated when the same operation is performed on the same component of two vectors with the same dimension or when each component of a vector is the same as the same scalar, these are basic vector operations
. In addition, an operation, such as concatenation, multiplication, or continuous comparison, can still be performed between each component of a vector to form a scalar. To improve the vector processing capability, basic vector operations can be performed
In order to have some flexibility, for example, to make certain components do not perform operations under the in-place vector control, or to add other special vector operations, for example, logical merge of two monotonic rising integer vectors with varying dimensions, compression and restoration of vectors.

Amount of streamline processing
Operations performed on each vector component are generally independent from each other, so they can be executed in parallel in multiple ways. This is vector-type parallel computing. Parallel Execution of vector operations, mainly using Pipeline
And array.

The size of the large-capacity high-speed primary memory (primary memory) capacity limits the machine's problem-solving scale. Vector Machines are mainly used to solve large-scale problems and must have a large capacity.
The primary storage of the data volume, and it should be a centralized public storage for convenient use and programming. When the High-speed computing pipeline starts, it is necessary to continuously supply the operands and take away the calculation results.
High data transmission rate. Otherwise, high-speed operations cannot be maintained.

The speed of the memory is always lower than that of the computing component. The data path between the memory and the computing component is a bottleneck that hinders the speed increase.
And the increase of the main storage capacity is in conflict with the increase of access speed. Therefore, it is one of the key issues in the design of vector machines to match the primary memory with the operation speed.

Vertical Machining Vector Machine

This type of machine uses the vertical Processing Method of the full length of the vector. Each execution of a vector operation must start from the beginning to the end to execute all the component operations. The operands or result vectors are directly taken from the primary memory or written to the primary memory. Data transmission rate of primary storage
3 ~ of the speed of the computing component ~ 4 times. The vertical processing vector machine sets a large number of storage bodies and a wide data path for cross-access, and accesses are performed in units of ultra-long words to meet the requirements. In this way,
High cost and low flexibility of the primary storage system, making it difficult to achieve efficient access to a wide range of primary storage vectors. In addition, the start time of vector operations is long, and the speed of short vector operations is greatly reduced.

Horizontal and horizontal Processing
Vector Machine
This type of machine uses the vector segmentation vertical and horizontal processing mode, and sets multiple vector operation registers with small capacity and high speed. When a Vector Operation expression is calculated, each vector operation only executes one segment of components at a time. Operations taken from primary storage
The number vector and the intermediate result vector generated by the operation can be stored in the vector register segment by segment. The operation component mainly accesses the vector register group. In this way, the computation component can perform high-speed computation and reduce the memory size.
To reduce the data transmission rate of primary storage by about 70% compared with vertical processing. Both CRAY-1 and 757 in China belong to this type.

Software and Application
Vector Machines are generally equipped with vector assembly and vector high-level languages for you to compile vector programs that can exert the specific vector machine's speed potential. Only vector-type parallel algorithms are developed and used to make the vectors included in the program
The more operations, the longer the vector, the higher the computing speed. Vector libraries for various application fields can be easily used by users and improve the efficiency of solving vector machines. The vector recognition program is newly developed in the Compilation Program.
This part is used to automatically identify the vector operation components in the source program using the usual Serial Algorithm during compilation, and compile it into the corresponding vector operation target program, to speed up Vector Machine computing of a large number of existing non-vector programs
Degree. Vector recognition technology needs to be further developed and improved to improve the recognition level.

The development direction of vector processing machine is multi-vector machine system or Cell Structure Vector Machine. The former must be implemented in software and computing
Progress has been made to solve many problems, such as task division and assignment. The latter must adopt an appropriate interconnection network, it is only possible to use hardware to automatically resolve conflicts arising from the coexistence and use of scattered primary storages as centralized.
Creates a virtual coexistence cellular structure vector machine. It not only has the advantages that the structure of the array machine is easy to expand the number of parallel sets to increase the speed, but also has the advantages that the vector machine is easy to use.

Contact Us

The content source of this page is from Internet, which doesn't represent Alibaba Cloud's opinion; products and services mentioned on that page don't have any relationship with Alibaba Cloud. If the content of the page makes you feel confusing, please write us an email, we will handle the problem within 5 days after receiving your email.

If you find any instances of plagiarism from the community, please send an email to: and provide relevant evidence. A staff member will contact you within 5 working days.

A Free Trial That Lets You Build Big!

Start building with 50+ products and up to 12 months usage for Elastic Compute Service

  • Sales Support

    1 on 1 presale consultation

  • After-Sales Support

    24/7 Technical Support 6 Free Tickets per Quarter Faster Response

  • Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.