On October 28–29, 2024, the VIRAC hosted a workshop as part of the RADIOBLOCS project, titled "New Science in Radio Astronomy: Applying Cutting-Edge Technology to Enhance the Entire Data Chain, from Receiver to Final Output." The workshop brought together participants from the VIRAC, ASTRON (Netherlands Institute for Radio Astronomy), and EPFL (École Polytechnique Fédérale de Lausanne).
The main objectives of the workshop were:
During the workshop, the use of potential GPU technologies for the project was discussed. Optimizing the use of computational resources requires selecting technologies that facilitate efficient communication between processes. The discussions explored several tools capable of improving algorithm performance while streamlining the software development process through optimized process management, eliminating the need for manual development of communication algorithms.
A key topic was the evaluation of the DASK framework, an open-source Python library designed for parallel and efficient processing of large datasets. DASK enables computations that surpass the memory or CPU limits of a single machine (see details: https://www.dask.org/). Its application is planned for the data processing needs of various interferometric telescope arrays, including the International LOFAR Telescope, the European Extremely Long Baseline Array, the Irbene Single Baseline Interferometer, and the upcoming Square Kilometre Array. The synergy between WP5.1 and other work packages of the RADIOBLOCKS project was also a key point of discussion.
The workshop further explored alternatives to the DASK framework, with a focus on Legate, a technology developed to seamlessly replace the popular Python package Numpy while leveraging GPU integration (see details: https://research.nvidia.com/publication/2019-11_legate-numpy-accelerated-and-distributed-array-computing). Employing multiple frameworks throughout the project could provide valuable insights and recommendations for developing next-generation multi-core software for radio astronomy. A comparative analysis of the technologies discussed during the workshop could serve as the foundation for a comprehensive publication, highlighting the performance of different technologies under various parameters.
These activities were conducted as part of the project "New Science in Radio Astronomy: Applying Cutting-Edge Technology to Enhance the Entire Data Chain, from Receiver to Final Output" (RADIOBLOCKS), funded under project No. 101093934.