Argonne National Laboratory

Swift provides scalable solution for APS beamline experiment

December 8, 2014

The Swift parallel scripting language, developed in the Mathematics and Computer Science Division, recently was used to enable efficient analysis of X-ray scattering data from the Advanced Photon Source on Mira, the IBM Blue Gene/Q supercomputer at Argonne National Laboratory.

Justin Wozniak, an assistant computer scientist in the Mathematics and Computer Science Division at Argonne, used Swift to leverage Mira’s computing power for scientific users of high-energy diffraction microscopy (HEDM), a technique for imaging the grain structures of polycrystalline materials. With Swift, Wozniak successfully transformed HEDM analysis software routines into a scalable application that was able to utilize Mira efficiently and return results to the APS user during the experiment.

Using Swift’s transformation of HEDM, Hemant Sharma, a postdoctoral researcher funded by General Electric Company, was able to analyze images collected from the Advanced Photon Source, detect and recalibrate the experiment for skewed data, and rerun multiple experiments – all in a far shorter time than previously possible.  This work will enhance the diagnosis of materials defects in high-stress applications such as jet engines, as well as other materials science investigations. 

For the full story of the collaboration, see “Boosting Beamline Performance” http://www.alcf.anl.gov/articles/boosting-beamline-performance.

For more information about Swift, see http://swift-lang.org

The work is presented here: http://www.mcs.anl.gov/publication/big-data-staging-mpi-io-interactive-x-ray-science