Data-centric software applications are becoming more difficult to engineer due to ever-increasing demands on computational platforms. Graphical Processing Units (GPUs) offer an alternative platform on which to deploy these applications. We propose a novel approach by which high-level, visual models of GPU applications are used to specify target applications and model-driven engineering techniques used to synthesize GPU code. We will invent new visual models, model-driven engineering techniques, and proof of concept support tools that will support verification via model checking and reuse of solutions across domains. We will validate our approach using example applications from molecular simulation and astrophysics domains.
|Effective start/end date||1/01/12 → 31/12/14|
- Australian Research Council (ARC): AUD109,053.40