A Memory Efficient Parallel All-Pairs Computation Framework: Computation - Communication Overlap

被引：1

作者：

Yeleswarapu, Venkata Kasi Viswanath ^{[1
]}

Somani, Arun K. ^{[1
]}

机构：

[1] Iowa State Univ, Dept Elect & Comp Engn, Ames, IA 50010 USA

来源：

PARALLEL PROCESSING AND APPLIED MATHEMATICS (PPAM 2017), PT I | 2018年 / 10777卷

基金：

美国国家科学基金会;

关键词：

Communication - computation overlap; High performance computing; All-Pairs problems; Parallel computing; MPI;

D O I：

10.1007/978-3-319-78024-5_39

中图分类号：

TP31 [计算机软件];

学科分类号：

081202 ; 0835 ;

摘要：

All-Pairs problems require each data element in a set of N data elements to be paired with every other data element for specific computation using the two data elements. Our framework aims to address recurring problems of scalability, distributing equal work load to all nodes and by reducing memory footprint. We reduce memory footprint of All-Pairs problems, by reducing memory requirement from N/root P to 3N/P. A bio-informatics application is implemented to demonstrate the scalability ranging up to 512 cores for the data set we experimented, redundancy management, and speed up performance of the framework.

引用

页码：443 / 458

页数：16

共 42 条

[41] FILMPAR: A parallel algorithm designed for the efficient and accurate computation of thin film flow on functional surfaces containing micro-structure
Lee, Y. C.
Thompson, H. M.
Gaskell, P. H.
COMPUTER PHYSICS COMMUNICATIONS, 2009, 180 (12) : 2634 - 2649
[42] Development of Efficient and Accurate Parallel Computation Algorithm Using Moving Overset Grids on Background Multi-Domains for Complex Two-Phase Flows
Ha, Minho
Cheong, Cheolung
Seol, Hanshin
Paik, Bu-Geun
Kim, Min-Jae
Jung, Young-Rae
APPLIED SCIENCES-BASEL, 2018, 8 (10):

← 1 2 3 4 5 →