GPUinspiral – a low-latency, high-performance implementation of the matched-filter gravitational wave search algorithm
Researh Institute for Particle and Nuclear Physics of the Hungarian Academy of Sciences, 1121, Budapest Konkoly-Thege Miklos u. 29-33 Hungary
RMKI Group, 2011
@article{debreczeni2011gpuinspiral,
title={GPUinspiral-a low-latency, high-performance implementation of the matched-filter gravitational wave search algorithm},
author={Debreczeni, G. and Somhegyi, B.},
year={2011}
}
A very high performance search pipeline has been developed for the search for gravitational wave signals originating from coalescing compact binary systems in the M<35 MSUN mass range. The goal of this research is to provide a solution to some of the so far computationally unfeasible data analysis methods such as for example the filtering of complete generic double spinning, precessing template banks using the matched-template filtering method, or the filtering of far much longer templates necessary to be used with the advanced gravitational wave detectors. In contrary to previous attempts [1] this implementation allows one to execute the entire analysis algorithms on many-core architectures such as GPUs, and not only a fraction of it which results in a drammatic acceleration of the search algorithm. In a typical inspiral pipeline the a speed-up of 92-fold has been achived using a Tesla C2050 GPU card respect to the reference compared with one core of 3.0 GHz AMD CPU. The features and the current performance of this pipeline are presented.
November 7, 2011 by hgpu
