Home // SIMUL 2015, The Seventh International Conference on Advances in System Simulation // View article

Memory Constrained Iterative Phase Retrieval using GPGPU

Authors:
Ladislav Mikeš

Keywords: Radiation scattering, Inverse problem, Reconstruction, GPGPU

Abstract:
Prediction of a scattering experiment for the new generation of coherent X-ray sources, such as X-Ray Free-Electron Laser (XFEL) requires a significant advance in both precision and numerical effectivity of radiation damage modeling. The preferred method of analyzing data from such experiment to obtain structure information relies on an iterative approach of refining an object estimate by repeatedly simulating the beam propagation between exiting the object and arriving at the detector. Currently, the most cost-effective way is to delegate work to General-Purpose computing on Graphics Processing Units (GPGPU), working around its limitations given by a reduced instruction set, limited memory size and increased latency when accessing the rest of the system. In this paper, we present managing fast GPU memory when performing reconstruction of an object consisting of up to 128 million voxels on currently available devices with less than 6GB of onboard memory.

Pages: 79 to 83

Copyright: Copyright (c) IARIA, 2015

Publication date: November 15, 2015

Published in: conference

ISSN: 2308-4537

ISBN: 978-1-61208-442-8

Location: Barcelona, Spain

Dates: from November 15, 2015 to November 20, 2015