Speaker
Description
The Mantid project is almost 20 years old. It started as a C++ GUI for processing muon and time-of-flight neutron scattering data at ISIS. In 2010, ORNL was the first of many to join in the development of the software. Several significant changes have occurred in scientific computing since the first commit to the Mantid repository in 2007. Among them is the rise of Python as a dominant language for scientific computing and the creation of several outstanding libraries. One example is that standard data objects (e.g. numpy arrays and dataframes) have become the lingua franca of scientific Python. This is increasingly relevant as Python becomes a language used for integrating and gluing together various functionalities.
At ORNL, data are measured in event mode, which captures attributes of every detected neutron. As a consequence, the significant power increase at the Spallation Neutron Source (SNS) (currently at 2.0MW) directly correlates with increases in data rates. Meantime, dramatic increases in the number of pixels per instrument correspondingly drive increased compute requirements. All of these factors are prompting a revisit of how Mantid is designed and implemented.
This talk will present the ORNL plan for modernizing Mantid in preparation for the increased power at SNS, event filtering at HFIR, with an eye towards the higher data rate and more pixelated instruments planned at the second target station.
