Fast period searches using the Lomb–Scargle algorithm on Graphics Processing Units for large datasets and real-time applications

Gowanlock, Michael; Kramer, Daniel; Trilling, David E.; Butler, Nathaniel R.; Donnelly, Brian

doi:10.1016/j.ascom.2021.100472

Cited by 9 publications

(10 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this section, we compare Su-perSmoother to Lomb-Scargle. We use the GPUaccelerated algorithm described in our prior work (Gowanlock et al, 2021). The GPU-accelerated Lomb-Scargle algorithm was shown to significantly outperform the parallel CPU algorithm, achieving speedups > 100× on several experimental scenarios (Gowanlock et al, 2021).…”

Section: Discussionmentioning

confidence: 99%

“…We use the GPUaccelerated algorithm described in our prior work (Gowanlock et al, 2021). The GPU-accelerated Lomb-Scargle algorithm was shown to significantly outperform the parallel CPU algorithm, achieving speedups > 100× on several experimental scenarios (Gowanlock et al, 2021). Table 6 shows the response time of deriving the periods using one GPU for the three datasets using the default grid search parameters in Table 2.…”

Section: Discussionmentioning

confidence: 99%

“…This will significantly reduce the cost of searching large frequency spaces with SuperSmoother. We can leverage our prior work in this area using our GPUaccelerated Lomb-Scargle algorithm (Gowanlock et al, 2021). This method would be similar to Shin and Byun (2004), which used a coarse-grained period search followed by a fine-grained search.…”

Section: Discussionmentioning

confidence: 99%

See 2 more Smart Citations

GPU-Enabled Searches for Periodic Signals of Unknown Shape

Gowanlock¹,

Butler²,

Trilling³

et al. 2021

Preprint

View full text Add to dashboard Cite

Recent and future generation observatories will enable the study of variable astronomical phenomena through their timedomain capabilities. High temporal fidelity will allow for unprecedented investigations into the nature of variable objects -those objects that vary in brightness over time. A major bottleneck in data processing pipelines is constructing light curve solutions for catalogs of variable objects, as it is well-known that period finding algorithms are computationally expensive. Furthermore, there are many period finding algorithms that are often suited for specific science cases. In this paper, we present the first GPU-accelerated Super Smoother algorithm. Super Smoother is general purpose and uses cross-validation to fit line segments to a time series, and as such, is more computationally expensive than other algorithms, such as Lomb-Scargle. Because the algorithm requires making several scans over the input time series for a tested frequency, we also propose a novel generalized-validation variant of Super Smoother that only requires a single scan over the data. We compare the performance of our algorithms to analogous parallel multi-core CPU implementations on three catalogs of data, and show that it is generally advantageous to use the GPU algorithm over the CPU counterparts. Furthermore, we demonstrate that our single-pass variant of Super Smoother is roughly equally as accurate at finding correct period solutions as the original algorithm. Our software supports several features, such as batching the computation to eliminate the possibility of exceeding global memory on the GPU, processing a single object or batches of objects, and we allow for scaling the algorithm across multiple GPUs.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

GPU-Enabled Searches for Periodic Signals of Unknown Shape

Gowanlock¹,

Butler²,

Trilling³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Almost all the 83 CPU hour effort is dedicated to LSP processing. To address this computational load, we have created a GPU LS implementation (Gowanlock et al 2021) that is described in Section 6.2.1. With this GPU approach, we can readily complete the daytime processing within one day, using our project-specific node on Monsoon that has 2 AMD EPYC 7542 CPUs (32 cores each) and 4 Nvidia A100 GPUs.…”

Section: Requirements and Current Performancementioning

confidence: 99%

“…Performance is critical for outlier detection tasks, as all-pairs searches have a worst-case quadratic time complexity. We are leveraging our prior work in this area to perform fast kNN searches, distance similarity searches, and clustering, which have been parallelized using GPUs, and has been published in the computer science literature (Gallet & Gowanlock 2019;Gowanlock 2019;Gowanlock & Karsin 2019;Gowanlock et al 2021).…”

mentioning

confidence: 99%

The Solar System Notification Alert Processing System (SNAPS): Design, Architecture, and First Data Release (SNAPShot1)

Trilling

Gowanlock

Kramer

et al. 2023

Self Cite

View full text Add to dashboard Cite

We present here the design, architecture, and first data release for the Solar System Notification Alert Processing System (SNAPS). SNAPS is a solar system broker that ingests alert data from all-sky surveys. At present, we ingest data from the Zwicky Transient Facility (ZTF) public survey, and we will ingest data from the forthcoming Legacy Survey of Space and Time (LSST) when it comes online. SNAPS is an official LSST downstream broker. In this paper we present the SNAPS design goals and requirements. We describe the details of our automatic pipeline processing in which the physical properties of asteroids are derived. We present SNAPShot1, our first data release, which contains 5,458,459 observations of 31,693 asteroids observed by ZTF from 2018 July to 2020 May. By comparing a number of derived properties for this ensemble to previously published results for overlapping objects we show that our automatic processing is highly reliable. We present a short list of science results, among many that will be enabled by our SNAPS catalog: (1) we demonstrate that there are no known asteroids with very short periods and high amplitudes, which clearly indicates that in general asteroids in the size range 0.3–20 km are strengthless; (2) we find no difference in the period distributions of Jupiter Trojan asteroids, implying that the L4 and L5 clouds have different shape distributions; and (3) we highlight several individual asteroids of interest. Finally, we describe future work for SNAPS and our ability to operate at LSST scale.

show abstract