This is an implementation of the Paper "Automatic Generation of Load-Balancing-Aware Block-Structured Grids for Complex Ocean Domains" presented at the International Meshing Roundtable 2022.

The provided example uses the ADCIRC mesh of the Bight of Abaco in the Bahamas region as an input. [Grenier Jr. et al.]

Requirements Linux:

(install in order as listed)

C++17

CMAKE

• cmake version ≥ 3.10
• follow installation guide: https://cmake.org/install/

CUDA

Requirements

• CUDA-capable GPU
• A supported version of Linux with a gcc compiler and toolchain
• NVIDIA CUDA Toolkit

Installation

• download CUDA
• To set path in .bashrc: get cuda installation path with

  <cuda_path>=ldconfig -p | grep cuda, e.g. /usr/local/cuda-11.0/  
  export CUDA_ROOT=<cuda_path>/bin/
  export LD_LIBRARY_PATH=<cuda_path>/lib64/
  

• Select nvidia graphics card with

  sudo prime-select nvidia
  

• reboot
• check if OpenGL uses nvidia GPU:

  glxinfo | grep -i opengl
  

Verify CUDA Installation

• Verify driver version:

  cat /proc/driver/nvidia/version
  

• Verify the CUDA Toolkit version:

  nvcc -V
  

• Verify running CUDA GPU jobs (Running deviceQuery and bandwidthTest)

  cuda-install-samples-[11.0].sh .
  cd NVIDIA_CUDA-[11.0]_Samples/
  make
  ./bin/[x86_64]/[linux]/release/deviceQuery
  ./bin/[x86_64]/[linux]/release/bandwidthTest
  

LIBS to install

1. glog
2. OpenMesh 8.1

• Create a directory for building the libraries (non-root should note the dependency)

  mkdir 00_libs
  cd 00_libs
  

• Clone the repositories and/or download the archives

  git clone https://github.com/google/glog.git
  wget https://www.graphics.rwth-aachen.de/media/openmesh_static/Releases/8.1/OpenMesh-8.1.tar.gz
  

• Build+Test glog (Root users may disregard the -DCMAKE_INSTALL_PREFIX flag)

  mkdir glog/build
  cd glog/build/
  cmake .. -G "Unix Makefiles" -DCMAKE_INSTALL_PREFIX="."
  cmake --build .
  cmake --build . --target test
  cmake --build . --target install
  cd ../..
  

• Build OpenMesh 8.1 (Root users may disregard the -DCMAKE_INSTALL_PREFIX flag)

  tar -xvf OpenMesh-8.1.tar.gz
  rm OpenMesh-8.1.tar.gz
  mkdir OpenMesh-8.1/build
  cd OpenMesh-8.1/build
  cmake .. -DCMAKE_INSTALL_PREFIX="."
  cmake --build .
  cmake --build . --target install
  cd ../..
  

• Note real path where LIBS have been installed (maybe add it to your bashrc extension if you plan on recompiling the project more often)

  realpath glog/build/ googletest/build/ OpenMesh-8.1/build/
  export TEMP_LIB_location=$(echo "$(realpath glog/build/);$(realpath OpenMesh-8.1/build)")
  

Build HPMeshGen2

• Clone the repository

  git clone https://github.com/DanielZint/hpmeshgen.git
  

• Building

  mkdir hpmeshgen2/build
  cd hpmeshgen2/build
  cmake .. -DCMAKE_PREFIX_PATH="$TEMP_LIB_location"
  cmake --build .
  cd ../..
  

• Executing with parameters described in ./hpmeshgen2/HPMeshGenParameter.txt

  ./hpmeshgen2/build/bin/HPMeshGen2
  

Requirements Windows:

(install in order as listed)

CUDA

• download
• normal installation, restart device afterwards

CMAKE

• download
• select binary with installer

VCPKG

• download
• follow along the video
• libs to install:
  1. glog
  2. OpenMesh 8.1

OPTIONAL

VISUAL STUDIO

• download
• select everything that has to do with C++ or MSVC
• select english language pack

Usage:

Use file HPMeshGenParameter.txt to configure HPMeshGen2. The file marks comments with '!'.

workingDirectory        ! is the project root folder by default but can be changed here
cacheFolder             ! storage folder for cached files (cache reduces time for recomputing meshes significantly)
nBlocks                 ! number of blocks

meshFileName            ! input mesh, usually a .14-file
sizegridSizeX           ! size grid dimensions (integer)
sizegridSizeY           ! the size grid stores the depth and edge length information

nRefinementSteps        ! the number of nodes on a fragment edge
nPatches                ! number of fragments

outputFolder            ! all output files go in here
initialMeshOutput       ! 0/1 initial mesh as .off file
reductionOutput         ! 0/1 intermediate steps of reduction as .off file
blockMeshOutput         ! 0/1 fragment mesh as .off file
sizegridOutput          ! 0/1 print size grid in .vtk format

forceFragmentNumber     ! 0/1 force fragment number, 0: change fragment number for better quality
convexHullDecimation    ! 0/1 activate masks