The OPS Domain Specific Abstraction for Multi-block Structured Grid Computations

Abstract

Code maintainability, performance portability and future proofing are some of the key challenges in this era of rapid change in High Performance Computing. Domain Specific Languages and Active Libraries address these challenges by focusing on a single application domain and providing a high-level programming approach, and then subsequently using domain knowledge to deliver high performance on various hardware.In this paper, we introduce the OPS high-level abstraction and active library aimed at multi-block structured grid computations, and discuss some of its key design points; we demonstrate how OPS can be embedded in C/C++ and the API made to look like a traditional library, and how through a combination of simple text manipulation and back-end logic we can enable execution on a diverse range of hardware using different parallel programming approaches.Relying on the access-execute description of the OPS abstraction, we introduce a number of automated execution techniques that enable distributed memory parallelization, optimization of communication patterns, checkpointing and cache-blocking. Using performance results from CloverLeaf from the Mantevo suite of benchmarks, we demonstrate the utility of OPS.

Implications of a metric for performance portability

“…Some have proposed the use of domain-specific languages (DSLs), providing a limited set of high-level abstractions for a specific domain alongside a mechanism for generating optimized platform-specific binaries (e.g. Liszt [7], OP2 [19], [25], [26], STELLA [10], Pochoir [31], PATUS [5], Halide [24] and Nabla [4]). Others have proposed portability frameworks that take a different approach, providing a lower-level abstraction of performance-enabling hardware for developers to program against: some of these tools are truly languages or language extensions (e.g.…”

Assessing the performance portability of modern parallel programming models using TeaLeaf

“…Herdman et al performed a similar analysis to the one presented in this research, evaluating OpenACC, OpenCL, and CUDA using a sister mini app of Tealeaf's, CloverLeaf. CloverLeaf has also been used to investigate the performance of the OPS DSL …”

“…CloverLeaf has also been used to investigate the performance of the OPS DSL. 26,27 Rupp et al 28 CPU with respect to microscopy image analysis. They concluded that the devices had a significant variance between particular operations, exposing some preference for particular operations.…”

Loop Tiling in Large-Scale Stencil Codes at Run-Time with OPS

“…The OPS (Oxford Parallel library for Structured meshes) DSL (Domain Specific Language) [9], [10] is essentially a C/C++/Fortran domain-specific API that uses source-tosource translation and various back-end libraries to automatically parallelise applications. Any code written using arXiv:1704.00693v2 [cs.PF] 26 Jun 2017 its API can utilise MPI, use multi-core CPUs with OpenMP, as well as GPUs with CUDA, OpenACC, or OpenCL.…”