streamit

StreamIt is a programming language and compiler framework designed for expressing streaming computations in the dataflow paradigm. It seeks to separate algorithmic specification from concerns about parallelism and hardware, making it easier to map streaming applications onto parallel architectures.

In StreamIt, programs are built as graphs of processing elements connected by channels that carry streams of data. Elements can be composed hierarchically, enabling complex pipelines to be built from simpler pieces. The language provides constructs to route data through the graph, allowing flexible streaming topologies and scalable parallel execution.

The execution model emphasizes scheduling and data movement. StreamIt includes tooling to perform static scheduling of the streaming graph for predictable throughput and resource usage, while supporting iterative refinement and experimentation for dynamic mapping scenarios. The associated compiler and runtime can target multiple backends, enabling exploration of mappings to multicore CPUs or specialized streaming hardware.

History and influence: StreamIt originated as an academic project in the early 2000s to study dataflow-based streaming programming. It has been used in university courses and research on compiler design, parallelization, and the hardware realization of streaming systems.

See also: Dataflow programming, stream processing, data-parallel languages.