Event generation at colliders such as the Large Hadron Collider is typically divided into two steps, integration and generation. Integration for high-multiplicity processes has involved weeks on a traditional cluster. This project will implement new Monte Carlo techniques for tackling the integration process more efficiently and for scaling such calculations on high-performance computing resources.
A challenge to scaling on existing Cray and IBM HPC resources is the variability in the perturbative calculations. To distribute the large intermediate results of these calculations, we will develop an event generation framework that is compatible with existing generators. This will improve scalability on DOE supercomputers.