One-stop strategy to search for long-duration gravitational-wave signals

TL;DR

This paper introduces fasttracks, a GPU-based engine that significantly improves the computational efficiency of searching for long-duration continuous gravitational-wave signals, enabling more effective and scalable data analysis.

Contribution

The paper presents fasttracks, a massively-parallel GPU engine for generic CW detection, and a scalable sensitivity estimation method, enhancing search efficiency without approximations.

Findings

Significant speedup in detection statistic evaluation using GPU parallelization.

Successful application to an all-sky search in LIGO-Virgo-KAGRA data.

Framework applicable to future long-duration gravitational-wave searches.

Abstract

Blind continuous gravitational-wave (CWs) searches are a significant computational challenge due to their long duration and weak amplitude of the involved signals. To cope with such problem, the community has developed a variety of data-analysis strategies which are usually tailored to specific CW searches; this prevents their applicability across the nowadays broad landscape of potential CW source. Also, their sensitivity is typically hard to model, and thus usually requires a significant computing investment. We present fasttracks, a massively-parallel engine to evaluate detection statistics for generic CW signals using GPU computing. We demonstrate a significant increase in computational efficiency by parallelizing the brute-force evaluation of detection statistics without using any computational approximations. Also, we introduce a simple and scalable postprocessing which allows us to formulate a generic semianalytic sensitivity estimate algorithm. These proposals are tested in a minimal all-sky search in data from the third observing run of the LIGO-Virgo-KAGRA Collaboration. The strategies discussed here will become increasingly relevant in the coming years as long-duration signals become a standard observation of future ground-based and space-borne detectors.