Expressing Performance and Redundancy in OpenMP: the LIONESS Annotations and Toolchain

Modern Systems on Chip (SoCs) incorporate multiple processing units, enabling various possibilities for parallelizing and accelerating computations. These processing capabilities can be leveraged to enhance resiliency by introducing temporal and spatial redundancy. In this paper, we propose a set of extensions to the OpenMP annotations that cover both performance and resiliency in a unique and homogeneous framework. We present the implementation of these extensions and their integration in a complete toolchain, and evaluate its capabilities on an image processing Use Case.