Abstract
Nowadays, some frameworks like Arcane and Dune offer a number of advanced tools to deal with the complexity related to parallelism, meshes and linear solvers. However, they do not handle the high level complexity related to discretization methods and physical models. Generative programming and Domain Specific Languages (DSL) are key technologies allowing to write code with a high level expressive language and take advantage of the efficiency of generated code with low level services. DSL may be embedded in host languages like Python or C++ . Such languages, named in that case Domain Specific Embedded Languages (DSEL), are applied for instance in frameworks like Fenics or Feel++ which are dedicated to the domain of Finite Element (FE) methods and Galerkin methods. ArcFVDSL is a DSEL developed on top of the Arcane framework, aiming to implement various lowest order methods (Finite-Volume (FV), Mimetic Finite Difference (MFD), Mixed Hybrid Finite Volume (MHFV), etc.) for diffusive problems on general meshes. In this paper, we present various implementations of different complex academic problems. We focus on the capability of the language to allow the description and the resolution of these problems with several lowest-order methods. We illustrate the benefits of such technology combined to runtime system tools like Heterogeneous Abstract RunTime System (HARTS) and its ability to handle seamlessly new heterogeneous architectures with multi-core processors enhanced by General Purpose computing on Graphics Processing Units (GP-GPU). We present the performance results of each implementation on different kinds of heterogeneous hardware architecture.
Highlights
Industrial simulation softwares have to manage: (i) the complexity of the underlying physical models, usually expressed in terms of a Partial Differential Equation (PDE) system completed with algebraic closure laws, (ii) the complexity of the numerical methods used to solve the PDE system, and (iii) the complexity of the low level computer science services required to have efficient software on modern hardware
In [3] we have presented ArcFVDSL, our Domain Specific Embedded Languages (DSEL) aiming to implement various lowest-order methods (Finite-Volume, Mimetic Finite Difference, Mixed Hybrid Finite Volume, etc.) for diffusive problems on general meshes
We focus on the capability of the language combined to runtime system tools like Heterogeneous Abstract RunTime System (HARTS) [5], to handle seamlessly new heterogeneous architectures with multi-core processors enhanced by General Purpose computing on Graphics Processing Units (GP-GPU)
Summary
Industrial simulation softwares have to manage: (i) the complexity of the underlying physical models, usually expressed in terms of a Partial Differential Equation (PDE) system completed with algebraic closure laws, (ii) the complexity of the numerical methods used to solve the PDE system, and (iii) the complexity of the low level computer science services required to have efficient software on modern hardware. This unified framework allows the design of a high level language close to the mathematical notation Such a language enables to express the variational discretized formulation of PDE problem with various methods, each of them defining specific bilinear and linear forms. It is important for the generative framework to have a high level unified way to express such sequences independently of low level optimisations We have for this purpose defined an abstract algebraic API detailed in Listing 3 aiming: (i) to hide hardware specificities, (ii) to manage memory allocation and locality, (iii) to manage parallel loops, (iv) to provide most BLAS 1 and 2 functionalities, (v) to provide tools to split vectors and manage vector views and range iterators.
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: Oil & Gas Science and Technology – Revue d’IFP Energies nouvelles
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
