| Swift | |
| Swift | |
| Paradigms: | Dataflow, distributed, grid, concurrent, scientific workflow, scripting |
| Designers: | --> |
| Developers: | University of Chicago, Argonne National Laboratory |
| Latest Release Version: | 0.96.2 |
| Typing: | Strong |
| Platform: | Cross-platform |
| Operating System: | Cross-platform |
| License: | Apache 2.0 |
| Influenced By: | C syntax, functional programming |
| Influenced: | Cuneiform |
Swift[1] is an implicitly parallel programming language that allows writing scripts that distribute program execution across distributed computing resources,[2] including clusters, clouds, grids, and supercomputers. Swift implementations are open-source software under the Apache License, version 2.0.
A Swift script[3] describes strongly typed data, application components, invocations of applications components, and the interrelations in the dataflow between those invocations. The program statements will automatically run in parallel unless there is a data dependency between them, given sufficient computing resources. The design of the language guarantees that results of a computation are deterministic, even though the order in which statements executes may vary. A special file data type is built into Swift. It allows command-line programs to be integrated into a program as typed functions. This allows programmers to write programs that treat command-line programs and files in the same way as regular functions and variables. A concept of mapping[4] is used to store and exchange complex data structures using a file system structure with files and directories.
Rapid dispatch of parallel tasks to a wide range of resources is implemented through a mechanism called Coasters task dispatch.[5] A Message Passing Interface based implementation of the language[6] supports very high task execution rates (e.g., 3000 tasks per second)[7] on large clusters and supercomputers.