HPC and GPU accelerated imaging toward the SKA era

Radio astronomy is currently facing a significant challenge due to the massive data volumes generated by modern radio-interferometers, which will be further exacerbated by the upcoming Square Kilometre Array. Efficient data processing at this scale necessitates advanced High Performance Computing (HPC) resources. Our work focuses on developing a novel approach to implement the w-stacking algorithm on state-of-the-art HPC systems, specifically targeting heterogeneous architectures comprising both CPUs and GPUs. We introduce the RICK (Radio Imaging Code Kernels) code, designed to efficiently process radio-interferometric data by leveraging the parallelism and computational power of modern HPC nodes. This study demonstrates the effectiveness of RICK on a single computing node, showcasing significant performance improvements over traditional methods. The paper outlines the methodology, the algorithmic innovations, and the parallelization strategy, along with performance benchmarks on various CPU/GPU configurations, highlighting the potential of RICK for future large-scale radio astronomy projects.