2023A&A...679A..20M

Context. Pulsar surveys with modern radio telescopes are becoming increasingly computationally demanding. This is particularly true for wide field-of-view pulsar surveys with radio interferometers and those conducted in real or quasi-real time. These demands result in data analysis bottlenecks that can limit the parameter space covered by the surveys and diminish their scientific return.
Aims. In this paper we address the computational challenge of 'candidate folding' in pulsar searching, presenting a novel, efficient approach designed to optimise the simultaneous folding of large numbers of pulsar candidates. We provide a complete folding pipeline appropriate for large-scale pulsar surveys that includes radio frequency interference mitigation, de-dispersion, folding, and parameter optimisation.
Methods. By leveraging the fast discrete dispersion measure transform (FDMT) algorithm, we have developed an optimised and cache-friendly implementation that we term the pruned FDMT (pFDMT). This implementation is specifically designed for candidate folding scenarios where the candidates are broadly distributed in dispersion measure space. The pFDMT approach efficiently reuses intermediate processing results and prunes the unused computation paths, resulting in a significant reduction in arithmetic operations. In addition, we propose a novel folding algorithm based on the Tikhonov-regularised least squares method that can improve the time resolution of the pulsar profile.
Results. We present the performance of its real-world application as an integral part of two major pulsar search projects conducted with the MeerKAT telescope: the MPIfR-MeerKAT Galactic Plane Survey (MMGPS) and the Transients and Pulsars with MeerKAT (TRAPUM) project. In our processing of approximately 500 candidates, the theoretical number of de-dispersion operations can be reduced by a factor of around 50 when compared to brute-force de-dispersion, which scales with the number of candidates.