Introduction to parallel processing

Only under the following conditions can some applications use parallel processing: there are enough applications that can make full use of multiprocessors; Parallelization target applications or users need to write new code to take advantage of parallel programs.

Traditionally, multiprocessors were specially designed for "parallel computers". According to this idea, Linux currently supports SMP Pentium system, in which multiple processors share a memory and bus interface in a computer. Each machine group running Linux can be connected with each other through the network to form a parallel processing group. The third option is to use the Linux system as the "host" and provide a special additional parallel processor. The fourth new option is SIMD parallelism in registers, which is suitable for multimedia extension (MMX).

The typical hardware environment for parallel processing is:

A single area on a single processor;

A single area in a multiprocessor (SMP);

Multi-zone configuration

Region in the processor (MPP);

Areas in a multiprocessor (SMP group);

Logical database area (also known as multi-logic code or MLN in DB2 PE, which is the DB2 parallel version of AIX version 1).

The possible types of parallelism in various hardware environments are as follows:

Internal parallel mechanism of I/O parallel query in hardware environment

Inter-partition parallelism of internal partitions

Single area, single processor? No?

Single-region, Multi-processor (SMP) is whether or not?

Is Multi-Region Single Processor (MPP) Yes or No?

Multi-region, Multi-processor (SMP group) Yes Yes Yes Yes Yes Yes

Logical database area Yes Yes Yes Yes Yes

Up to now, the representative application fields of parallel computer are: weather forecasting modeling, computer-aided design of VLSI, large-scale database management, artificial intelligence, crime control and national defense strategy research, and its application scope is also expanding. Parallel processing technology is mainly a structured technology with algorithm as the core, parallel language as the description and software and hardware as the implementation tools, which are interrelated and mutually restricted.