MODESTO: Data-centric Analytic Optimization of Complex Stencil Programs on Heterogeneous Architectures

Code transformations, such as loop tiling and loop fusion, are of key importance for the efficient implementation of stencil compu-tations. However, their direct application to a large code base is costly and severely impacts program maintainability. While re-cently introduced domain-specific languages facilitate the appli-cation of such transformations, they typically still require manual tuning or auto-tuning techniques to select the transformations that yield optimal performance. In this paper, we introduce MODESTO, a model-driven stencil optimization framework, that for a stencil program suggests program transformations optimized for a given target architecture. Initially, we review and categorize data lo-cality transformations for stencil programs and introduce a sten-cil algebra that allows the expression and enumeration of different stencil program implementation variants. Combining this algebra with a compile-time performance model, we show how to auto-matically tune stencil programs. We use our framework to model the STELLA library and optimize kernels used by the COSMO atmospheric model on multi-core and hybrid CPU-GPU architec-tures. Compared to naive and expert-tuned variants, the automat-ically tuned kernels attain a 2.0–3.1x and a 1.0–1.8x speedup re-spectively