A dataflow-inspired CGRA for streaming applications

The herein presented research is motivated by the need for reconfigurable, flexible computing arrays targeted at streaming applications that contain a large degree of instruction-level parallelism. Such arrays are usually referred to as coarse-grained reconfigurable arrays (CGRAs). CGRAs are composed of small, reconfigurable cores that are interconnected to form a computing grid. When reviewing related work, it became clear to us that an intuitive, easy-to-use programming paradigm was missing. The vast majority of the available CGRAs use a C-based programming approach. In our opinion however, C is not the most obvious choice to program such systems, since C has originally been designed as a sequential language, and therefore lacks support for structural description of parallelism. As a consequence, the burden of extracting the structure of an implemented algorithm lies on the compiler. Here, we present a complete CGRA, consisting of an architecture and a programming language. Both the archi- tecture and the programming language are inspired by the principles found in dataflow. In particular, the firing rule, which means that a node fires as soon as all the required inputs are available. Furthermore, instead of developing a programming paradigm based on C, we decided to base our programming paradigm on the base of the functional programming language Haskell, that already has a notion of structure and the ability to describe instruction-level parallelism.