Multi-scale CLEAN Memory Utilisation Challenges at Square Kilometre Array Scale

Deconvolution is a very important tool in radio astronomy, used to remove the point spread function from the image produced by an interferometer. Currently popular deconvolution algorithms are up to 50 years old and were designed for much smaller instruments than are currently being constructed. At image sizes proposed for the Square Kilometre Array (SKA), in this paper we show that the commonly used msCLEAN algorithm developed by Cornwell requires up to 13 TB of memory to successfully run. We detail the steps of the msCLEAN algorithm and the incremental memory usage for each. We also propose some memory reduction techniques to improve performance.