Analysis of neutron sensitivity and data-flow error detection in ARM microprocessors using NEON SIMD extensions

摘要

This work analyzes the sensitivity to neutrons of SIMD (Single Instruction Multiple Data) microprocessor extensions. To this purpose, the ARM SIMD coprocessor (NEON (TM)) was selected as a case study and neutron radiation experiments were performed on a commercial device running NEON software. In addition, we analyze the benefits of using the NEON coprocessor as a means to efficiently implement data-flow hardening approaches. Experimental results show that SIMD extensions have a great potential to improve performance and reduce the overheads associated to software data-flow hardening.