Assembling a High-Productivity DSL for Computational Fluid Dynamics

Abstract

As we move towards exascale computing, an abstraction for effective parallel computation is increasingly needed to overcome the maintainability and portability of scientific applications while ensuring the efficient and full exploitation of high-performance systems. These circumstances require computer and domain scientists to work jointly toward a productive working environment. Domain specific languages address this challenge by abstracting the highlevel application layer from the final, complex parallel low-level code. Saiph is an innovative domain specific language designed to reduce the work of computational fluid dynamics domain experts to an unambiguous and straightforward transcription of their problem equations. The high-level language, domain-specific compiler and underlying library are enhanced to make applications developed by scientists intuitive. Additions and improvements are presented, designed for the significant advantage of running computational fluid dynamics applications on different machines with no porting or maintenance issues. Numerical methods and parallel strategies are independently added at the library level covering the explicit finite differences resolution of a vast range of problems. Depending on the application, a specific parallel resolution is automatically derived and applied within Saiph, freeing the user from decisions related to numerical methods or parallel executions while ensuring suitable computations. Through a list of benchmarks, we demonstrate the utility and productivity of the Saiph high-level language together with the correctness and performance of the underlying parallel numerical algorithms. CCS CONCEPTS• Software and its engineering → Domain specific languages; • Computing methodologies → Parallel computing methodologies; • Applied computing → Physical sciences and engineering.

Automated Generation of High-Performance Computational Fluid Dynamics Codes

“…We use several Saiph applications [2] to evaluate our contributions. These applications are open-source [5], and they include a reference output to validate their correctness.…”

“…Results are presented in the last column of Table 1. Saiph based on lambdas outperforms old implementations [1,2] by a factor of 4x to 14x depending on the length of the equation trees since the new implementation avoid traversing it at each time step. This performance increase comes from the fact that lambda functions generated at compile-time enable native compiler optimisations that cannot be applied when tree-traversals take place at runtime.…”

“…Saiph is a DSL that targets the resolution of Computational Fluid Dynamics (CFD) problems through a high-level syntax and a generic numerical library implementing explicit and Finite Differences Methods (FDM). We use Saiph [1,2] as a DSL platform and consequently focus on the CFD domain. However, the modular design of the tool enables the present methodologies to be generalised for other problem domains according to their computational needs.…”

Combining One-Sided Communications with Task-Based Programming Models

“…Lastly, we want to mention that our Task-Aware GASPI library is being used in Saiph [25] [26], which is a Domain-Specific Language (DSL) that facilitates the simulation of physical phenomena from the Computational Fluid Dynam-ics (CFD) domain in HPC systems. Macià et al [27] extend Saiph with a new back-end that generates a hybrid taskbased GASPI+OmpSs-2 variant of a high-level application and internally leverages the TAGASPI library to perform one-sided communications.…”