ESTEST is a framework to facilitate the verification and comparison of Electronic structure codes like Qbox and Quantum Espresso. These community codes are developed by disparate teams and institutions but compute some of the same fundamental physical quantities. Quantitative and qualitative comparisons between these different implementations impart a degree of certainty to new calculations (as compared to established results) against human and machine errors. Additionally, a group of related calculations may be evaluated based on their agreement with each other and their common physical foundation. An individual researcher then, may gain confidence through such a presentation of peer-reviewed information, and the access to that knowledge becomes a tool useful to the community.
Automated Code Execution LayerThe base of ESTEST automates the collection of physical simulation results from input files to electronic structure codes using QMTest by the methods of regression testing.
Translation to Unified Representation XML DatabaseOutput garnered from the execution of an electronic structure code is then translated to a common XML data format that is a unified representation (UR). For each electronic structure code there exists a script to perform this translation written in Python.
Web Accessible DatabaseUR XML is then stored in a Xquery database that forms the foundation for a rich web portal powering the dynamic post-processing, validation, and comparison of physical simulation results.
At this time all three technical components are implemented at an alpha state of development incorporating the Quantum Espresso and Qbox electronic structure codes. A working demonstation has been shown at the APS March meeting.
Work remains to be done to open this framework in beta. Groups and individuals are welcome to test this framework with their own distribution and in particular translate their own electronic structure code results. This distributed model of having many sites running a local ESTEST server and producing their own expert data raises questions on how to synchronize and share data between these sources.
Gary Yuan [ email@example.com ]
Francois Gygi [ fgygi at ucdavis dot edu ]