Pulscan: Binary Pulsar Detection Using Unmatched Filters on NVIDIA GPUs

The Fourier Domain Acceleration Search (FDAS) and Fourier Domain Jerk Search(FDJS) are proven matched filtering techniques for detecting binary pulsarsignatures in time-domain radio astronomy datasets. Next generation radiotelescopes such as the SPOTLIGHT project at the GMRT produce data at rates thatmandate real-time processing, as storage of the entire captured dataset forsubsequent offline processing is infeasible. The computational demands of FDASand FDJS make them challenging to implement in real-time detection pipelines,requiring costly high performance computing facilities. To address this wepropose Pulscan, an unmatched filtering approach which achievesorder-of-magnitude improvements in runtime performance compared to FDAS whilstbeing able to detect both accelerated and some jerked binary pulsars. Weprofile the sensitivity of Pulscan using a distribution (N = 10,955) ofsynthetic binary pulsars and compare its performance with FDAS and FDJS. Ourimplementation of Pulscan includes an OpenMP version for multicore CPUacceleration, a version for heterogeneous CPU/GPU environments such as NVIDIAGrace Hopper, and a fully optimized NVIDIA GPU implementation for integrationinto an AstroAccelerate pipeline, which will be deployed in the SPOTLIGHTproject at the GMRT. Our results demonstrate that unmatched filtering inPulscan can serve as an efficient data reduction step, prioritizing datasetsfor further analysis and focusing human and subsequent computational resourceson likely binary pulsar signatures.