Beyond von Neumann: weakly programmable processor arrays and their programming

In English The age of parallelism is here. For a sustainable software development for massively parallel architectures the von Neumann model need to be replaced by one with native support for parallelism. We suggest a data flow model for signal processing applications. This will make it possible to reuse software implementations for different targets and future platform generations. We also outline our development tool flow for compiling CAL, a data flow language, to parallel architectures. We also present our processor array, which can be configured to handle massively parallel computations. We demonstrate its power by implementing part of a software radio receiver.