CppCon 2015

Today one of the most important barriers between performing an experiment and publishing the results is the complex and time-consuming effort that individual researchers apply to data reduction and analysis. In a neutron experiment one must relate measured counts to a physically meaningful four-dimensional scattering function. While the first neutron scatters had to manually setup their instrument and record neutron counts – in handwriting – one single point at a time, users at a modern time of flight spectrometer rapidly collect hundreds of gigabytes of data in less than one day. High performance software is now a necessity. Rapid reduction, analysis and visualization provide feedback for users, helping them make optimal use of limited beam time. Software must also be accessible to and customized by users with a wide range of abilities and interests in programming. In this talk, I will discuss Mantid-(the Manipulation and Analysis Toolkit for Instrument Data), which is the framework designed for the reduction and analysis of data acquired at muon and neutron sources. It is a partnership between multiple user facilities including Rutherford Appleton Laboratory, Oxfordshire, UK, Oak Ridge National Laboratory, Tennessee, USA and the European Spallation Source, Scandinavia. The core of the Mantid Framework is written in C++ with the user interacting via a GUI or python bindings. I will describe real-world cross-platform situations we have encountered while supporting Mantid on Linux, OS X and Windows. This plus a climate that values stability over the latest operating systems and compilers creates challenges to utilizing new language and library features available in C++11/14. To better exploit parallelism on all of the supported platforms, we are wrapping parallel loops so that different implementations can be chosen at compile time. Considerable portions of Mantid provide domain-specific user interfaces to cross-platform libraries. One example is 3D visualization of the large multidimensional datasets generated in Mantid. Our approach utilizes the application widgets and rendering libraries from ParaView. Mantid provides plugins specifically designed for the Mantid data structures. A custom interface for visualization can be started within Mantid, creating a seamless environment for novice users not comfortable with the ParaView workflow.

This research at ORNL’s High Flux Isotope Reactor and Spallation Neutron Source was sponsored by the Scientific User Facilities Division, Office of Basic Energy Sciences, US Department of Energy.