michaelst98 / miluphpc

High Performance Computing Smooth(ed) Particle Hydrodynamics

---

Michael Staneker

---

Some ...

languages

              

databases

     

operating systems

                 

and other stuff

        

I already worked with and like.

Further experience with , programming in . Moreover, hands-on experience with HPC and parallel computing for shared-memory systems distributed-memory systems and GPUs as well as with parallel file I/O via . Furthermore, I did some web development for which I already used , , , , and .

---

Projects

MilupHPC

milupHPC API documentation A HPC N-Body and Smoothed Particle Hydrodyamics (SPH) code via CUDA-aware MPI targeting distributed memory systems For 1, 2 and 3 dimensional simulation Dynamic load balancing using space-filling curves Current test cases: Plummer model Taylor–von Neumann–Sedov blast wave Isothermal collapse of a molecular cloud (Boss & Bodenheimer)

More information ...

This repository aims to implement a Multi-GPU SPH/NBody algorithm using CUDA aware MPI by combining ideas from:

• Single-GPU version inspired/adopted from:

  • Miluphcuda
  • An Efficient CUDA Implementation of the Tree-Based Barnes Hut n-Body Algorithm
  • Implementation: MichaelSt98/NNS CUDA_NBody

• Multi-Node (or rather Multi-CPU) version inspired/adopted from:

  • M. Griebel, S. Knapek, and G. Zumbusch. Numerical Simulation in Molecular Dynamics: Numerics, Algorithms, Parallelization, Applications. 1st. Springer Pub- lishing Company, Incorporated, 2010. isbn: 3642087760
  • Implementation: MichaelSt98/NNS (branch: MolecularDynamics)

• some more samples: each color represents a process, thus a GPU

• Kepler disk

  • Kepler disk: four GPUs (hilbert curve)

• Plummer model
  • four GPUs with dynamic load balancing every 10th step (top: lebesgue, bottom: hilbert)

• Taylor–von Neumann–Sedov blast wave
  • Sedov explosion: one and two GPUs

• Boss-Bodenheimer: isothermal collapse
  • one and two GPUs

---

miluphpc's Issues

MPIOFileDriver from HighFive is deprecated since v2.5.0

Compatibility problem with HighFive since v2.5.0.
Further information in this issue

Proposed Solution:

// open a new file with the MPI IO driver for parallel Read/Write
File file(FILE_NAME,
    File::ReadWrite | File::Create | File::Truncate,
    FileAccessProps(MPI_COMM_WORLD, MPI_INFO_NULL));

instead of

MPIOFileDriver(MPI_COMM_WORLD, MPI_INFO_NULL);