Adámek Karel

Abstract Reference: 30835
Identifier: P3.1
Presentation: Poster presentation
Key Theme: 3 New Trends in HPC and Distributed Computing 

A real-time Single Pulse detection algorithm for GPUs  

Authors:
Adámek Karel, Armour Wes

Detecting non-repeating events in the radio spectrum has become increasingly important in radio astronomy over the last decade due to the discovery of fast radio bursts (FRBs). Because we do not know the signal properties a priori, we choose to use a series of boxcar filters of differing lengths. The boxcar filter matching the single pulse width of a signal present in the data sweeps up the signal lifting it out of the background noise. Given the vast data rates of next generation radio telescopes and the number of computational operations required by boxcar filtering, to achieve real-time detection the use of High Performance Computing techniques and hardware is required. We have implemented a GPU single pulse detection algorithm, for NVIDIA GPUs, which use boxcar filters of varying widths. Our code performs thresholding based on the signal-to-noise ratio produced by a boxcar filter of a given width and presents the highest signal-to-noise ratio detected in the data. We present our parallel implementation of our single pulse detection algorithm, along with performance results for SKA like data.