SIMD acceleration of a TLM solver using ARM processors

In recent years, investigations into accelerating Transmission Line Matrix (TLM) solver programs have concentrated on the use of large-scale parallel computing systems. The focus for such systems rests on leveraging multi-core software architecture principles to achieve improved performances. Mirroring desktop CPU advances of the early 2000s, modern small-scale computing platforms are increasingly transitioning to dual or even quad-core CPU architectures to provide developers with opportunities for improving software performance. This paper discusses an alternative to traditional multi-core techniques for obtaining performance increases with such architectures. Using the NEON Single Instruction, Multiple Data (SIMD) instruction set present in the ARMv7 architecture processors often found within modern small-scale computing devices, we demonstrate performance increases nearly two times base figures without the normal pitfalls of multi-core software design.