Performance Models for Data Transfers: A Case Study with Molecular Chemistry Kernels - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2019

Performance Models for Data Transfers: A Case Study with Molecular Chemistry Kernels

Abstract

In distributed memory systems, it is paramount to develop strategies to overlap the data transfers between memory nodes with the computations in order to exploit their full potential. In this paper, we consider the problem of determining the order of data transfers between two memory nodes for a set of independent tasks with the objective of minimizing the makespan. We prove that, with limited memory capacity, the problem of obtaining the optimal data transfer order is NP-complete. We propose several heuristics to determine this order and discuss the conditions that might be favorable to different heuristics. We analyze our heuristics on traces obtained by running two molecular chemistry kernels, namely, Hartree-Fock (HF) and Coupled Cluster Singles Doubles (CCSD), on 10 nodes of an HPC system. Our results show that some of our heuristics achieve significant overlap for moderate memory capacities and resulting in makespans that are very close to the lower bound.
Fichier principal
Vignette du fichier
dataTransferModels.pdf (4.24 Mo) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-02431877 , version 1 (08-01-2020)

Identifiers

Cite

Suraj Kumar, Lionel Eyraud-Dubois, Sriram Krishnamoorthy. Performance Models for Data Transfers: A Case Study with Molecular Chemistry Kernels. ICPP 2019 - 48th International Conference on Parallel Processing, Aug 2019, Kyoto, Japan. ⟨10.1145/3337821.3337921⟩. ⟨hal-02431877⟩

Collections

CNRS INRIA INRIA2
72 View
93 Download

Altmetric

Share

Gmail Facebook X LinkedIn More