A PRESTO-based parallel pulsar search pipeline used for FAST drift scan data

Abstract

We developed a pulsar search pipeline based on PulsaR Exploration and Search TOolkit (PRESTO). This pipeline simply runs dedispersion, Fast Fourier Transform (FFT) and acceleration search in process-level parallel to shorten the processing time.With two parallel strategies, the pipeline can highly shorten the processing time in both normal searches and acceleration searches. This pipeline was first tested with Parkes Multibeam Pulsar Survery (PMPS) data and discovered two new faint pulsars. Then, it was successfully applied in processing the Five-hundred-meterAperture Spherical radio Telescope (FAST) drift scan data with tens of new pulsar discoveries up to now. The pipeline is only CPU-based and can be easily and quickly deployed in computing nodes for testing purposes or data processing.

Arecibo and FAST Timing Follow-up of twelve Millisecond Pulsars Discovered in Commensal Radio Astronomy FAST Survey

“…The equivalent width W (s), S/N are estimated from the summed pulse profile. PSR J0203−0150 is an MSP with a spin period of 5.17 ms and was first reported in Yu et al (2020). It is in a 49.96 day binary orbit with an eccentricity of 0.0001938.…”

Searching for pulsars with phase characteristics

“…With such a large amount of intermediate data and complex search processing, it is a time-consuming process to get those pulsar candidates. As an example, it takes 170 second to process a data set of 26 second observation with a 24-cores computing (Yu et al 2020). To this end, a bunch of accelerated schemes have been provided, but most of them are carried out under the existing framework, such as using GPU to accelerate on the dedispersion (Barsdell et al 2012) and acceleration search (e.g., PRESTO 1 (PulsaR Exploration and Search TOolkit) on gpu 2 by Luo et al in prep), using FPGA as a backend for pulsar dedispersion processing (Luo et al 2017), or using computer clusters to parallelize the entire search (Yu et al 2020).…”

“…As an example, it takes 170 second to process a data set of 26 second observation with a 24-cores computing (Yu et al 2020). To this end, a bunch of accelerated schemes have been provided, but most of them are carried out under the existing framework, such as using GPU to accelerate on the dedispersion (Barsdell et al 2012) and acceleration search (e.g., PRESTO 1 (PulsaR Exploration and Search TOolkit) on gpu 2 by Luo et al in prep), using FPGA as a backend for pulsar dedispersion processing (Luo et al 2017), or using computer clusters to parallelize the entire search (Yu et al 2020). These methods have a relatively significant improvement in computing speed, but there are some problems in their application, such as how to deal with the memory computing bottleneck (Sclocco et al 2016) when using a single machine for dedispersion.…”

Searching for Pulsars with Phase Characteristics

“…With such a large amount of intermediate data and complex search processing, it is a timeconsuming process to get those pulsar candidates. As an example, it takes 170 s to process a data set of 26 s observation with a 24 cores computing (Yu et al 2020). To this end, a bunch of accelerated schemes have been provided, but most of them are carried out under the existing framework, such as using GPU to accelerate on the dedispersion (Barsdell et al 2012) and acceleration search (e.g., PRESTO 7 (PulsaR Exploration and Search TOolkit) on gpu 8 by Luo et al 2022, in preparation), using FPGA as a backend for pulsar dedispersion processing (Luo et al 2017), or using computer clusters to parallelize the entire search (Yu et al 2020).…”

“…As an example, it takes 170 s to process a data set of 26 s observation with a 24 cores computing (Yu et al 2020). To this end, a bunch of accelerated schemes have been provided, but most of them are carried out under the existing framework, such as using GPU to accelerate on the dedispersion (Barsdell et al 2012) and acceleration search (e.g., PRESTO 7 (PulsaR Exploration and Search TOolkit) on gpu 8 by Luo et al 2022, in preparation), using FPGA as a backend for pulsar dedispersion processing (Luo et al 2017), or using computer clusters to parallelize the entire search (Yu et al 2020). These methods have a relatively significant improvement in computing speed, but there are some problems in their application, such as how to deal with the memory computing bottleneck (Sclocco et al 2016) when using a single machine for dedispersion.…”