A framework for replica exchange simulations

Replica Exchange (RE) simulations have emerged as an important algorithmic tool for the molecular sciences. Typically RE functionality is integrated into the molecular simulation soft- ware package. A primary motivation for the tight integration of RE functionality with sim- ulation codes has been performance. This is limiting at multiple levels. First, advances in the RE methodology are tied to the molecular simulation code for which they were developed. Second, it is difficult to extend or experiment with various RE methods, since expertise in the molecular simulation code is required. We propose the RepEx framework which addresses the aforementioned limitations, while striking the balance between flexibility (RE pattern) and scalability (several thousand replicas) over a diverse range of HPC platforms. In this thesis is introduced the RepEx framework, the primary contributions of which are: (i) its ability to support different RE patterns independent of molecular simulation codes, (ii) the ability to execute different exchange types and replica counts independent of the specific availability of resources, (iii) a runtime system that has first-class support for task- level parallelism, and (iv) provide the required scalability along multiple dimensions.M.S.Includes bibliographical referencesby Antons Treikali