The Swift parallel scripting language enables scientists, engineers, and data analysts to express and coordinate parallel invocations of application programs on distributed and parallel computing platforms: one of the dominant modes of performing computation in science and engineering. Swift runs on a variety of HPC clusters and clouds, and enables users to move their application scripts between computing environments with relative ease.
Swift comprises a programming model, scripting language, and runtime engine. Its implicitly parallel programming model allows users with minimal programming expertise to transparently utilize parallel and distributed systems. The scripting language is simple, minimal and standalone; the programming model has also been embedded into the Python and R languages.
Swift is employed in many diverse domains, including biochemistry, neuroscience, climate model analysis, earthquake simulation, hydrology, energy forecasting, economics modeling, mass media analysis, materials science, and astronomy.
This project makes usability and programmatic expressiveness enhancements to Swift, broadens its utility library, performs the testing and hardening needed to serve a large-scale national and global community, and extends existing documentation and training material in a manner that will create and serve a broad user base. It also develops enhancements in configuration, script debugging, exception handling, parallel collection processing, and data typing and mapping to make Swift increasingly productive.
Swift enables users with little or no experience in parallel programming to leverage parallel and distributed computing environments ranging in scale from desktops to petascale supercomputers. It opens up powerful cyberinfrastructure like XSEDE, Open Science Grid, FutureGrid, and Blue Waters to a wide range and scale of scientific user communities, thus broadening participation in high performance computing.
