A dynamically customizable process-centered evaluation model

Abstract

We present the graph operational semantics approach used for defining NiMo nets execution, which mix lazy, data-driven and a weak form of eager evaluation, all in parallel. NiMo is a totally graphic language from the family of higher order typed languages but with a strong data-flow inspiration. Programs are process networks that evolve showing the full state at each execution step and can be dynamically changed or completed. In NiMo parallelization is implicit. The conjunction of two kinds of tags determine which processes are able to act in the same execution step. According to their tag, processes can behave in five modes that can be globally or locally set for each one, and can be also dynamically changed. Combining modes gives a very flexible way to experiment different strategies to increase processor usage, decrease channel population, and achieve subnet synchronization. Together with symbolic execution it also provides the means for generative and multi-stage programming.

Graphical and Incremental Type Inference: A Graph Transformation Approach

“…But we think it was worth the try; the graphic-functional-dataflow characteristics of NiMo result in a very powerful computation model where everything can be dynamically changed, even the evaluation policy. The NiMo execution model is described in [12]. We are currently working on several aspects of net visualization, which are critical when nets grow, such as the non-expanded view of net processes.…”