Michael A. Bauer, Alain Biem, et al.
HPC/SpringSim 2011
Synchrotrons can be used to generate X-rays in order to probe materials at the atomic level. One approach is to use X-ray diffraction (XRD) to do this. The data from an XRD experiment consists of a sequence of digital image files which for a single scan could consist of hundreds or even thousands of digital images. Existing analysis software processes these images individually sequentially and is usually used after the experiment is completed. The results from an XRD detector can be thought of as a sequence of images, generated during the scan by the X-ray beam. If these images could be analyzed in near real-time, the results could be sent to the researcher running the experiment and used to improve the overall experimental process and results. In this paper, we report on a stream processing application to remove background from XRD images using a pipelining implementation. We describe our implementation techniques of using IBM Infosphere Streams for parsing XRD source data and removing the background. We present experimental results showing the super-linear speedup attained over a purely sequential version of the algorithm on a quad-core machine. These results demonstrate the potential of making good use of multi-cores for high-performance stream processing of XRD images. © 2010 IOP Publishing Ltd.
Michael A. Bauer, Alain Biem, et al.
HPC/SpringSim 2011
Alain Biem, Nathan Caswell
HICSS 2008
Michael A. Bauer, Alain Biem, et al.
HPC/SpringSim 2011
Roongroj Nopsuwanchai, Alain Biem
ICASSP 2003