5G multi-antenna algorithm implementation and performance analysis for ARM Neon & SVE2 architectures
Loading...
URL
Journal Title
Journal ISSN
Volume Title
School of Electrical Engineering |
Master's thesis
Authors
Date
Department
Major/Subject
Mcode
Language
en
Pages
70
Series
Abstract
With the increasing demand for high computation power in modern communication technologies, finding alternatives to the traditional processors is a critical challenge nowadays. This thesis will propose using ARM Centralized Processing Units (CPUs) with the latest released Single Instruction Multiple Data (SIMD) acceleration architectures, which prioritize parallel data element processing and became applicable for communication systems. To evaluate the performance of the CPUs, a very commonly used Minimum Mean Square Error (MMSE) algorithm was targeted. To understand CPUs capability improvements over the years, two of the recently published SIMD architectures were compared in terms of accuracy and execution duration. Simulations and performance measurements of the implementations on target architectures showed that ARM CPUs are able to complete complicated complex matrix operations with accuracy. On the other hand, the computation speed of the CPUs is not close to the traditional physical layer processors, such as Digital Signal Processors (DSPs). Furthermore, to understand the evolution of CPUs throughout the time, a comparison of the earlier architecture NEON and one of the recent architectures Signal Vector Extension 2 (SVE2), implies that there is no big performance enhancement, and even in specific operations, the earlier technique, NEON, performs better. In conclusion, even though ARM CPUs can be used either in low load requiring scenarios or by multiplying CPU cores in more complex cases, there is still room for improvement to compete with regular signal processing machines.Description
Supervisor
Wichman, RistoThesis advisor
Mathecken, Pramod JacobAiroldi, Roberto