ABSTRACT

Stream programs represent an important class of high-performance computing. These programs are rich in parallelism and can be naturally targeted to distributed and multi-core architectures. Since computer architectures gradually become multi-core, there is a pressing need to provide an efficient programming language that supports all aspects of parallelism in the streaming applications. In this paper, we introduce a new flexible stream programming language, called FSPL. The FSPL, is an architecture-independent programming language designed for high-performance streaming application development. It aims to improve programmer productivity and program efficiency within the streaming domain. In the FSPL language, each program is a collection of independent filters which communicate by the means of data channels. This model lends itself naturally to concurrent and efficient implementations on modern multiprocessors. One of most significant features supported in FSPL is that when you define a filter, it is not needed to specify the amount of data produced and consumed by that filter.