Skip to content

Latest commit

 

History

History
158 lines (120 loc) · 8.13 KB

runpetibm.md

File metadata and controls

158 lines (120 loc) · 8.13 KB
Raw

Run PetIBM

PetIBM is a library (a toolbox to solve the Navier-Stokes equations with an immersed-boundary method) that comes with several application codes. Upon successful installation, the library (shared and/or static) and the application programs should be respectively located in the lib and bin folders of your installation directory. For example, if installing PetIBM from Anaconda channel barbagroup, then the installation directory is the root of the Conda environment, which is $CONDA_PREFIX.

Once PetIBM is installed, the libraries (shared and/or static) are located in the lib folder of your installation directory. The present software package comes with 5 application codes that use the PetIBM library. Upon successful installation, the binary executables for these applications are located in the bin folder of your installation directory. If PetIBM is not installed using conda (or mamba), you can prepend the PATH environment variable with the bin directory to use the binary executables:

$ export PATH=<petibm-installation-directory>/bin:$PATH

List of binary executables: * petibm-navierstokes * petibm-ibpm * petibm-decoupledibpm * petibm-writemesh * petibm-vorticity * petibm-createxdmf

These programs work for 2D and 3D configurations and can be run in serial or parallel (with mpiexec or mpirun). The following sub-sections provide more details about each executable.

Program petibm-navierstokes

The program solves the 2D or 3D Navier-Stokes equations using a projection method: an approximate block-LU decomposition (Perot, 1993). The equations are fully discretized in space (second-order central differences) and in time. At each time step, the program solves a system for the velocity field and then a Poisson system for the pressure to project the velocity onto the space of divergence-free fields.

To run the program:

$ cd <simulation-directory>
$ petibm-navierstokes

In the current working directory differs from the simulation directory, provide the path of the simulation directory:

$ petibm-navierstokes -directory <simulation-directory>

If the YAML configuration file is not located in the simulation directory:

$ petibm-navierstokes -config <config-path>

To run on two CPU processes:

$ mpiexec -np 2 petibm-navierstokes

Program petibm-ibpm

The program solves the 2D and 3D Navier-Stokes equations with an immersed-boundary method: the Immersed Boundary Projection Method (IBPM) proposed by Taira and Colonius (2007). The method is based on the projection approach of Perot (1993) and is suitable to solve the flow around one or several, fixed or moving, rigid immersed boundaries. Note that the current implementation of the method in PetIBM only works for fixed immersed boundaries. At each time step, the program solves a system for the velocity field and then a modified Poisson system for the pressure field and Lagrangian forces. Finally, the velocity field is projected onto the space of divergence-free fields that satisfy the no-slip condition at the location of the Lagrangian boundary points.

To run a simulation with the IBPM:

$ cd <simulation-directory>
$ mpiexec -np n petibm-ibpm

You can also provide the path of the simulation directory with the command-line argument -directory <path> and/or the path of the YAML configuration file with -config <path>.

Program petibm-decoupledibpm

The program solves the 2D and 3D Navier-Stokes equations with an immersed-boundary method: a decoupled version of the Immersed Boundary Projection Method proposed by Li et al. (2016). The method extends from the work of Taira and Colonius (2007) and decouples the pressure field from the Lagrangian boundary forces. Three linear systems are solved every time step: a system for the velocity field, a Poisson system for the pressure field, and a system for the Lagrangian forces. The divergence-free and the no-slip constraints are imposed in a sequential fashion.

To run a simulation with a decoupled version of the IBPM:

$ cd <simulation-directory>
$ mpiexec -np n petibm-decoupledibpm

You can also provide the path of the simulation directory with the command-line argument -directory <path> and/or the path of the YAML configuration file with -config <path>.

Program petibm-writemesh

This program is a simple (and optional) pre-processing utility that creates a structured Cartesian mesh based on the configuration provided in a given YAML file. The gridline coordinates are then written into a HDF5 file (grid.h5) saved in the simulation directory.

Program petibm-vorticity

This program is a post-processing utility to compute the vorticity vector field from the velocity vector field; it works for 2D and 3D configurations.

In 2D, petibm-vorticity:

  • reads the grids for the x- and y-components of the velocity field (group names u and v in the file grid.h5),
  • computes the grid of the vorticity scalar field,
  • writes the vorticity grid in grid.h5 (group name wz),
  • for each saved time step:
    • reads the x- and y-components of the velocity field (group names u and v in the HDF5 solution file),
    • computes the vorticity scalar field,
    • writes the vorticity values in the HDF5 solution field (group name wz).

In 3D, petibm-vorticity:

  • reads the grids for the x-, y- and z-components of the velocity field (group names u, v, and w in the file grid.h5),
  • computes the grid for each component of the vorticity vector field,
  • writes the vorticity grids in grid.h5 (group name wx, wy, and wz),
  • for each saved time step:
    • reads the x-, y-, and z-components of the velocity field (group names u, v, and w in the HDF5 solution file),
    • computes the vorticity vector field,
    • writes the vorticity values in the HDF5 solution field (group names wx, wy, and wz).

To compute the vorticity field:

$ cd <simulation-directory>
$ mpiexec -np n petibm-vorticity

You can also provide the path of the simulation directory with the command-line argument -directory <path> and/or the path of the YAML configuration file with -config <path>.

Program petibm-createxdmf

This program is a post-processing utility which creates XDMF files for the solution variables which can be used to visualize and process HDF5 fields with VisIt (versions 2.12.3 and 2.13.1 have been tested) or ParaView (versions 5.4.1 and 5.5.0 have been tested). The utility works for 2D and 3D configurations.

To create XDMF files:

$ petibm-createxdmf

It will create XDMF files for the pressure (p.xmf), the velocity components (u.xmf and v.xmf for 2D configurations;u.xmf, v.xmf, and w.xmf for 3D configurations), and the vorticity components (wz.xmf for 2D configurations; wx.xmf, wy.xmf, and wz.xmf for 3D configurations).

Running PetIBM using NVIDIA AmgX

To solve one or several linear systems on CUDA-capable GPU devices, PetIBM calls the NVIDIA AmgX library. Solving on GPU devices is supported in the programs petibm-navierstokes, petibm-ibpm, petibm-decoupledibpm.

For example to run a simulation using 2 MPI processes and 2 GPU devices, simply use:

$ mpiexec -np 2 <petibm-executable>

Or, if the MPI was installed using conda/mamba, you may need to use

$ mpiexec --mca opal_cuda_support 1 -np 2 <petibm-executable>

If you choose to run on 4 MPI processes but only using 2 GPU devices:

$ export CUDA_VISIBLE_DEVICES=0,1
$ mpiexec -np 4 <petibm-executable>

Again, --mca opal_cuda_support 1 may be needed if using Anaconda's OpenMPI package.

References

  • Li, R. Y., Xie, C. M., Huang, W. X., & Xu, C. X. (2016). An efficient immersed boundary projection method for flow over complex/moving boundaries. Computers & Fluids, 140, 122-135.
  • Perot, J. B. (1993). An analysis of the fractional step method. Journal of Computational Physics, 108(1), 51-58.
  • Taira, K., & Colonius, T. (2007). The immersed boundary method: a projection approach. Journal of Computational Physics, 225(2), 2118-2137.