From physics model to results: An optimizing framework for cross-architecture code generation

Starting from a high-level problem description in terms of partial differential equations using ab-stract tensor notation, the Chemora framework dis-cretizes, optimizes, and generates complete high performance codes for a wide range of compute architectures. Chemora extends the capabilities of Cactus, facilitating the usage of large-scale CPU/GPU systems in an efficient manner for com-plex applications, without low-level code tuning. Chemora achieves parallelism through MPI and multi-threading, combining OpenMP and CUDA. Optimizations include high-level code transforma-tions, efficient loop traversal strategies, dynam-ically selected data and instruction cache usage strategies, and JIT compilation of GPU code tai-lored to the problem characteristics. The dis-cretization is based on higher-order finite differ-ences on multi-block domains. Chemora’s capabil-ities are demonstrated by simulations of black hole collisions. This problem provides an acid test of the framework, as the Einstein equations contai