GeST: Generalized Stencil Auto-tuning Framework on GPUs

Stencil computations are widely used in high performance computing (HPC) applications. In recent years, stencils have become more diverse in terms of stencil order, memory accesses, and computation patterns. To adapt diverse stencils to GPUs, a variety of optimization techniques have been proposed. Due to the diversity of stencil patterns and GPU architectures, no single optimization technique fits all stencils. Therefore, stencil auto-tuning mechanisms have been proposed to conduct parameter searches for the combination of optimization techniques. However, existing mechanisms introduce large offline overhead and are inflexible to generalize to arbitrary stencil patterns. We propose GeST, a generalized auto-tuning framework that efficiently determines the optimal parameter setting of the global optimization space for stencils on GPUs. The experimental results show that GeST can identify better-performing settings in a short time compared to the state-of-the-art works.