Distributed stacking data storage method supporting SIMD system structure

Abstract

The invention discloses a distributed stacking data storage method supporting an SIMD system structure. Stacking spaces are allocated in an internal storage in a distribution mode, scalar stacks storing scalar information are allocated in a scalar storage, and vector stacks storing vector information are allocated in a vector storage; when a program is compiled, local variables needing to be accessed by scalar units are allocated in the scalar stacks, and local variables needing to be accessed by vector units are allocated in the vector stacks; when the program is operated, the scalar information, needing to be stored, in a program switching site, is stored in the scalar stacks, and vector information, needing to be stored, in a program switching site, is stored in the scalar stacks, and when the program returns on site, the scalar information is directly read from the scalar stacks to the scalar units, and the vector information is directly read from the vector stacks to the vector units. The distributed stacking data storage method supporting the SIMD system structure has the advantages of being high in storing and accessing speed of stacking data, small in bandwidth requirement, high in system performance and low in power consumption.

Description

technical field [0001] The invention relates to the field of SIMD processors, in particular to a distributed stack data storage method supporting SIMD architecture. Background technique [0002] With the increase of the scale of the problem and the improvement of real-time requirements, the processing power of the microprocessor with only the scalar unit has been difficult to meet the application requirements. The SIMD (Single Instruction Data, single instruction multiple data stream) vector processor is especially The unit's microprocessor is widely used in the industry. [0003] The running state of the program is generally managed by the compiler through the stack, that is, a block of memory is allocated in the memory as a stack at runtime, and the context, local variables, and dynamically allocated data areas of the function running process are managed to support function calls, interrupts, etc. processing and allocation of dynamic data areas, etc. At present, the cons...

AI Technical Summary

Problems solved by technology

Assuming that the SIMD microprocessor contains P vector units, a vector parameter of a 32-bit integer type has a length of 32*P bits, so reading vector data from the stack placed in the scalar memory takes a long time to access and Occupies a large bandwidth;

[0006] (2) Due to the large bit width of the vector data, it takes a long time to occupy the access port of the scalar memory when storing and accessing the stack, thus blocking the memory access of the scalar unit;

For example, when reading vector data from the stack, it is necessary to transfer the stack data from the scalar memory to the vector memory through DMA (Direct Memory Access), and then read the data from the vector memory to the vector register through the vector memory access operation. , so accessing vector data requires two steps, which further reduces system performance

Images