PaScaL_TDMA  v2.0
Documentation
Loading...
Searching...
No Matches
Installation

Table of Contents

Downloads

The repository can be cloned as follows:

git clone https://github.com/MPMC-Lab/PaScaL_TDMA.git

Alternatively, the source files can be downloaded through github menu 'Download ZIP'.

Compile

Prerequisites

Prerequisites to compile PaScaL_TDMA are as follows:

  • MPI (OpenMPI, IntelMPI, MPICH, .. and so on)
  • Fortran compiler (NVIDIA HPC SDK, Intel fortran, GNU fortran, .. and so on )
    • (nvfortran for GPU runs, NVIDIA HPC SDK 21.1 or higher)

Option1: NVIDIA HPC SDK + Open MPI

  • The NVIDIA HPC SDK includes a pre-compiled version of OpenMPI.

  • Change 24.3 to Your SDK version

    cd <your-path>/PaScaL_TDMA
    export TDMA_PATH=$(pwd)
    export NVARCH=`uname -s`_`uname -m`
    export PATH=/opt/nvidia/hpc_sdk/$NVARCH/24.3/compilers/bin:$PATH
    export cuda_version=$(nvcc --version | grep "release" | sed 's/.*release //' | sed 's/,//' | cut -d' ' -f1); export cuda_version
    export MPI=/opt/nvidia/hpc_sdk/$NVARCH/24.3/comm_libs/$cuda_version/openmpi4/latest
    export PATH=$PATH:/opt/nvidia/hpc_sdk/$NVARCH/24.3/compilers/bin
    export opt=module

  • Add shortcuts (Optional)

    # Also, change `24.3` to `Your SDK version`
    echo 'export NVARCH=`uname -s`_`uname -m`' >> ~/.bashrc
    echo 'alias nvmpirun="/opt/nvidia/hpc_sdk/$NVARCH/24.3/comm_libs/mpi/bin/mpirun"' >> ~/.bashrc
    source ~/.bashrc

  • Build PaScaL_TDMA

    make all
  • Test examples
    cd examples
    nvmpirun -np 2 ex1_single.out

Option2: Intel Fortran + Open MPI

  • Install Intel oneAPI Base Toolkit and Intel HPC Toolkit

  • Install Open MPI

    cd ~/PaScaL_TDMA
    export TDMA_PATH=$(pwd)
    # Or use 'source /opt/intel/oneapi/<version>/oneapi-vars.sh'
    source /opt/intel/oneapi/setvars.sh
    export MPI=<your-path>/openmpi
    export opt=J

  • Build PaScaL_TDMA

    make all
  • Test examples
    cd examples
    mpirun -np 2 ./ex1_single.out

Option3: Intel Fortran + Intel MPI

  • Install Intel oneAPI Base Toolkit and Intel HPC Toolkit

    cd ~/PaScaL_TDMA
    export TDMA_PATH=$(pwd)
    # Or use 'source /opt/intel/oneapi/<version>/oneapi-vars.sh'
    source /opt/intel/oneapi/setvars.sh
    export opt=module

  • Build PaScaL_TDMA

    make all
  • Test examples
    cd examples
    mpirun -np 2 ./ex1_single.out

Compile and build

  • Build PaScaL_TDMA
    make lib

  • Build an example problem after build PaScaL_TDMA

    make example

  • Build all

    make all
  • Clean build
    export TDMA_PATH=$(pwd)
    make clean

Mores on compile option

The Makefile in root directory is to compile the source code, and is expected to work for most systems. The 'Makefile.inc' file in the root directory can be used to change the compiler (and MPI wrapper) and a few pre-defined compile options depending on compiler, execution environment and et al.

Running the example

After building the example file, an executable binary, *.out, is built in the run folder. The PARA_INPUT.inp file in the run folder is a pre-defined input file, and the *.out can be executed as follows:

mpirun -np 8 ./a.out ./PARA_INPUT.inp

GPU power monitoring

In the tool folder, there is a Python script gpu_power_monitor.py that can be used to monitor and print real-time GPU power usage with timestamps. To use this script, you will need to install the pynvml library.

Folder structure

  • src : source files of PaScaL_TDMA 2.0.
  • example : source files of an example problem for 3D heat-transfer equation.
  • include : header files are created after building
  • lib : a static library of PaScaL_TDMA 2.0 is are created after building
  • doc : documentation
  • run : an executable binary file for the example problem is created after building.
  • tool : contains useful scripts and tools.
Previous Next
Introduction Performance