This title appears in the Scientific Report :
2011
Please use the identifier:
http://hdl.handle.net/2128/4453 in citations.
A priori minimisation of algorithmic bottlenecks in the parallel tree code PEPC
A priori minimisation of algorithmic bottlenecks in the parallel tree code PEPC
The challenging problems arising from fast parallel N-body simulations became a driver for high performance computing. The Barnes-Hut tree code is an example in the class of fast summation algorithms, with a complexity of O(N log(N)), instead of O(N2). The multi disciplinary code Pepc – the ’Pretty...
Saved in:
Personal Name(s): | Hübner, H. (Corresponding author) |
---|---|
Contributing Institute: |
Jülich Supercomputing Center; JSC |
Imprint: |
Jülich
Forschungszentrum Jülich GmbH Zentralbibliothek, Verlag
2011
|
Physical Description: |
102 p. |
Dissertation Note: |
Aachen, FH, Campus Jülich, Masterarbeit, 2011 |
Document Type: |
Master Thesis |
Research Program: |
Pretty Efficient Parallel Coulomb Solver Computational Science and Mathematical Methods Scientific Computing |
Series Title: |
Berichte des Forschungszentrums Jülich
4339 |
Subject (ZB): | |
Link: |
OpenAccess |
Publikationsportal JuSER |
The challenging problems arising from fast parallel N-body simulations became a driver
for high performance computing. The Barnes-Hut tree code is an example in the class of
fast summation algorithms, with a complexity of O(N log(N)), instead of O(N2). The
multi disciplinary code Pepc – the ’Pretty Efficient Parallel Coulomb solver’ – is based
on the Hashed-Oct-Tree scheme and is developed at Juelich Supercomputing Centre.
The pure bookkeeping overhead of the data-distributed tree construction decreases the
performance and rapidly increases for large scales, as shown for JUGENE, an IBM Blue
Gene/P architecture. For this reason, novel approaches will be established and applied,
minimising integral bottlenecks. An axiomatic and provable optimisation of the parallel
organisation structure, induced by a distributed memory machine, is introduced in detail.
Reducing memory footprint and communication alike, the new concept intrinsically guides
to a tight a-priori estimation of parallel data overhead. Moreover, the influence of the
locality-preserving Hilbert-curve on the irregular communication structure, is studied.
Accordingly, the new method provides an immense upgrade for the particle number,
making Pepc a more versatile tool for simulations in a multi disciplinary context |