TY  - CONF
AU  - Iwainsky, Christian
AU  - Shudler, Sergei
AU  - Calotoiu, Alexandru
AU  - Strube, Alexandre
AU  - Knobloch, Michael
AU  - Bischof, Christian
AU  - Wolf, Felix
TI  - How Many Threads will be too Many? On the Scalability of OpenMP Implementations
VL  - 9233
CY  - Berlin, Heidelberg
PB  - Springer Berlin Heidelberg
M1  - FZJ-2016-00659
SN  - 978-3-662-48095-3 (print)
T2  - Lecture Notes in Computer Science
SP  - 451 - 463
PY  - 2015
AB  - Exascale systems will exhibit much higher degrees of parallelism both in terms of the number of nodes and the number of cores per node. OpenMP is a widely used standard for exploiting parallelism on the level of individual nodes. Although successfully used on today’s systems, it is unclear how well OpenMP implementations will scale to much higher numbers of threads. In this work, we apply automated performance modeling to examine the scalability of OpenMP constructs across different compilers and platforms. We ran tests on Intel Xeon multi-board, Intel Xeon Phi, and Blue Gene with compilers from GNU, IBM, Intel, and PGI. The resulting models reveal a number of scalability issues in implementations of OpenMP constructs and show unexpected differences between compilers.
T2  - 21st International Conference on Parallel and Distributed Computing
CY  - 24 Aug 2015 - 28 Aug 2015, Vienna (Austria)
Y2  - 24 Aug 2015 - 28 Aug 2015
M2  - Vienna, Austria
LB  - PUB:(DE-HGF)8
UR  - <Go to ISI:>//WOS:000363786800035
DO  - DOI:10.1007/978-3-662-48096-0_35
UR  - https://juser.fz-juelich.de/record/280948
ER  -