Large-scale MD simulation of heterogeneous systems with ls1 mardyn - - PowerPoint PPT Presentation

▶
large scale md simulation of heterogeneous systems with
SMART_READER_LITE
LIVE PREVIEW

Large-scale MD simulation of heterogeneous systems with ls1 mardyn - - PowerPoint PPT Presentation

Dec 23, 2022 183 likes •348 views

Laboratory of Engineering Thermodynamics (LTD) Prof. Dr.-Ing. H. Hasse Large-scale MD simulation of heterogeneous systems with ls1 mardyn M. T. Horsch, R. Srivastava, S. J. Werth, C. Niethammer, C. W. Glass, W. Eckhardt, A. Heinecke, N.

slide-1
SLIDE 1

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

Large-scale MD simulation of heterogeneous systems with ls1 mardyn

  • M. T. Horsch, R. Srivastava, S. J. Werth, C. Niethammer, C. W. Glass,
  • W. Eckhardt, A. Heinecke, N. Tchipev, H.-J. Bungartz, S. Eckelsbach,
  • J. Vrabec and H. Hasse

TU Kaiserslautern, Engineering Thermodynamics (LTD) TU München, Scientific Computing in Computer Science (SCCS) High Performance Computing Centre Stuttgart (HLRS) University of Paderborn, Thermodynamics and Energy Technology (ThEt)

Frankfurt am Main, 23rd March 2015 ProcessNet International Workshop MolMod

slide-2
SLIDE 2

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

2

Parallelization by volume decomposition

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse

Linked-cell data structure suitable for spatial domain decomposition:

(non-blocking, overlap- ping MPI send/receive

  • perations)

large systems “1”: molecular dynamics http://www.ls1-mardyn.de/

slide-3
SLIDE 3

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

3

Parallelization by volume decomposition

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse

Methods for heterogeneous

  • r fluctuating particle

distributions: Linked-cell data structure suitable for spatial domain decomposition:

(non-blocking, overlap- ping MPI send/receive

  • perations)
slide-4
SLIDE 4

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

4

Scale separation and long-range correction

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse

Long-range correction from the density profile, following Janeček. For planar interfaces: Full evaluation of all pairwise interactions is too expensive ... ... instead, short-range interactions are evaluated for neighbours. short range (explicit) long range (correction) cutoff radius

slide-5
SLIDE 5

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

5

Scale separation and long-range correction

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse

Long-range correction from the density profile, following Janeček. Angle-averaging expression for multi-site models, following Lustig. For planar interfaces: Full evaluation of all pairwise interactions is too expensive ... ... instead, short-range interactions are evaluated for neighbours. short range (explicit) long range (correction) cutoff radius

slide-6
SLIDE 6

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

6

Molecular simulation of fluids at interfaces

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse

Long-range correction from the density profile, following Janeček. For arbitrary geometries, e.g. the fast multipole method can be employed. Two-centre LJ fluid (2CLJ) Janeček-Lustig term no angle averaging no correction at all Angle-averaging expression for multi-site models, following Lustig. For planar interfaces: 1 nm cutoff radius / σ surface tension / εσ -2

slide-7
SLIDE 7

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

7

Molecular simulation of fluids at interfaces

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse

2CLJQ models:

  • 2 LJ centres
  • Quadrupole

Test of predictivity for interfacial properties Model validation and optimization

slide-8
SLIDE 8

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

8

Molecular simulation of fluids at interfaces

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse

  • Adsorption (fluid-fluid and fluid-solid)
  • Vapour-liquid surface tension
  • Curved vapour-liquid interfaces
  • Contact angle and contact line pinning

LJTS T = 0.8 ε θpl = 90°

slide-9
SLIDE 9

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

9

MD simulation of nanofluidics

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse

slide-10
SLIDE 10

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

10

Scale bridging from nano- to microfluidics

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse

slide-11
SLIDE 11

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

11

Scaling of ls1 mardyn on hermit

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse

slide-12
SLIDE 12

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

12

Scaling of ls1 mardyn on hermit

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse

homogeneous cavitation

CO2 (T = 280 K and ρ = 17.2 mol/l), 3CLJQ 25 million molecules on 110 592 cores

slide-13
SLIDE 13

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

13

Optimization of ls1 mardyn for SuperMUC

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse

SuperMUC (LRZ Garching): 3 PFLOPS Intel Xeon Sandy Bridge cluster. Efficient vectorization:

  • Optimization by hand, using advanced vector extensions (AVX).
  • Conversion from array of structures (AoS) to structure of arrays (SoA).

Discussed in detail by Nikola Tchipev tomorrow. hyperthreaded sliding window forces acting on molecules are only stored while the cell is inside the sliding window

slide-14
SLIDE 14

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

14

Large-scale MD simulations on SuperMUC

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse

Scaling of ls1 mardyn examined on up to 146 016 cores, i.e. the whole SuperMUC, by Wolfgang Eckhardt and Alexander Heinecke in 2013. i d e a l s t r

  • n

g s c a l i n g

  • bserved strong scaling

number of cores speedup (relative to 128 cores) homogeneous LJTS liquid with 4.8 billion molecules

slide-15
SLIDE 15

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

15

Large-scale MD simulations on SuperMUC

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse

Up to N = 4 · 1012 molecules on SuperMUC number of cores speedup

2013

weak scaling with 31.5 million molecules per core

slide-16
SLIDE 16

Laboratory of Engineering Thermodynamics (LTD)

  • Prof. Dr.-Ing. H. Hasse

Free registration for ls1 mardyn at http://www.ls1-mardyn.de/ released as Free Software (BSD license)

16

Release of ls1 mardyn

Horsch, Srivastava, Werth, Niethammer, Glass, Eckhardt, Heinecke, Tchipev, Bungartz, Eckelsbach, Vrabec and Hasse