A Low Level Component Model enabling Resource Specialization of HPC Applications

Julien Bigot 1 Zhengxiong Hou 1 Christian Pérez 1, * Vincent Pichon 1, 2
* Corresponding author
1 AVALON - Algorithms and Software Architectures for Distributed and HPC Platforms
Inria Grenoble - Rhône-Alpes, LIP - Laboratoire de l'Informatique du Parallélisme
Abstract : Scientific applications are still getting more complex, e.g. to improve their accuracy by taking into account more phenomena. Moreover, computing infrastructures are continuing their fast evolution. Therefore, software engineering is becoming a major issue to achieve easiness of development, portability, simple maintenance, while achieving high performance. Software component model is a promising approach, which enables to manipulate the software architecture of an application. However, existing models do not capture enough resource specificities. This paper proposes a low level component model (L2C) that supports directly native connectors such as MPI, shared memory and method invocation. L2C is intended to be used as a back end by a ''compiler'' (such as HLCM) to generate an application assembly specific to a given machine. This paper shows on a typical domain decomposition use case that \llc can achieve the same performance as native implementations, while gaining benefits such as enabling resource specialization capabilities.
Complete list of metadatas

Cited literature [18 references]  Display  Hide  Download

https://hal.inria.fr/hal-00698573
Contributor : Christian Perez <>
Submitted on : Wednesday, May 16, 2012 - 6:09:28 PM
Last modification on : Tuesday, July 10, 2018 - 2:28:02 PM
Long-term archiving on : Thursday, December 15, 2016 - 7:34:18 AM

File

RR-7966.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-00698573, version 1

Collections

Citation

Julien Bigot, Zhengxiong Hou, Christian Pérez, Vincent Pichon. A Low Level Component Model enabling Resource Specialization of HPC Applications. [Research Report] RR-7966, INRIA. 2012, pp.20. ⟨hal-00698573⟩

Share

Metrics

Record views

483

Files downloads

346