PARCOACH Extension for a Full-Interprocedural Collectives Verification - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2018

PARCOACH Extension for a Full-Interprocedural Collectives Verification

Abstract

The advent to exascale requires more scalable and efficient techniques to help developers to locate, analyze and correct errors in parallel applications. PARallel COntrol flow Anomaly CHecker (PARCOACH) is a framework that detects the origin of collective errors in applications using MPI and/or OpenMP. In MPI, such errors include collective operations mismatches. In OpenMP, a collective error can be a barrier not called by all tasks in a team. In this paper, we present an extension of PARCOACH which improves its collective errors detection. We show our analysis is more precise and accurate than the previous one on different benchmarks and real applications.
Fichier principal
Vignette du fichier
correctness_2018.pdf (331.25 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-01937316 , version 1 (28-11-2018)

Identifiers

  • HAL Id : hal-01937316 , version 1

Cite

Pierre Huchant, Emmanuelle Saillard, Denis Barthou, Hugo Brunie, Patrick Carribault. PARCOACH Extension for a Full-Interprocedural Collectives Verification. Second International Workshop on Software Correctness for HPC Applications, Nov 2018, Dallas, United States. ⟨hal-01937316⟩
134 View
228 Download

Share

Gmail Facebook X LinkedIn More