Vector single instruction multiple data-stream (SIMD) operation structure supporting synergistic working of scalar and vector

Abstract

The invention discloses a vector single instruction multiple data-stream (SIMD) operation structure supporting synergistic working of a scalar and a vector. The vector SIMD operation structure comprises an united fetch and instruction issuing part, a scalar processing unit SPU, a vector processing unit VPU, a vector array storage AM and a direct memory access (DMA) unit, wherein united fetch and instruction issuing part is used for simultaneously issuing instructions to the scalar processing unit SPU, the vector processing unit VPU and the vector array storage AM, the scalar processing unit SPU is used for processing a serial task and executing control on the vector processing unit VPU, the vector processing unit VPU is used for processing a parallel task with intensive computation, the vector array storage AM is used for proving data and migration support for parallel and multi-width vector operation, and the DMA unit is used for providing the instructions and data for the scalar processing unit SPU and the vector processing unit VPU. By the vector SIMD operation structure, the execution efficiency and the parallelism can be comprehensively improved.

Description

technical field [0001] The invention mainly relates to the field of microprocessor structure and design, in particular to a vector SIMD operation structure supporting scalar-vector cooperative work. Background technique [0002] Digital signal processor (Digital Signal Processor, DSP) is widely used in embedded systems as a typical embedded microprocessor. The development of signal processing has brought great opportunities, and its application fields have expanded to all aspects of military and economic development. In the application fields of modern communication, image processing and radar signal processing, as the amount of data processing increases, the requirements for calculation accuracy and real-time performance increase, and it is usually necessary to use a higher-performance microprocessor for processing. [0003] Different from the traditional CPU, DSP has the following characteristics: (1) Strong computing power, focusing on real-time computing rather than con...

AI Technical Summary

Problems solved by technology

[0004] General scientific computing requires high-performance DSP, but traditional DSP has the following disadvantages when used in scientific computing: (1) The bit width is small, which makes the calculation accuracy and addressing space insufficient

However, when GPDSP uses many processing units to develop general scientific computing parallelism, there are still several problems as follows: (1) How to organize many isomorphic processing units so that it can efficiently develop multi-level parallelism in general scientific computing; (2) ) How to effectively coordinate the scalar operation unit used for control and the vector operation unit used for calculation; (3) How to provide support for matrix operations in general scientific computing, and use the large amount of data reuse characteristics in matrix operations to improve The data supply capability of many isomorphic processing units, thereby improving the computing efficiency of GPDSP

Images