Versions Compared

Old Version 1

changes.mady.by.user Chompoonut Rugnim

Saved on

compared with

New Version 2

changes.mady.by.user Chompoonut Rugnim

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Here are the steps to install Amber20 at TARA HPC.

  1. Obtain Amber20 at http://ambermd.org and download source codes

  2. Upload and extract amber source codes at your project home director

    Code Block
    cd /tarafs/data/project/projxxxx  #change projxxxx to your project ID  
tar xvfj AmberTools20.tar.bz2  
tar xvfj Amber20.tar.bz2

  3. Set the AMBERHOME home directory environment variable

    Code Block
    cd amber20_src
export AMBERHOME=$PWD

  4. Load modules and prepare for AMBER20 installation

    Code Block
    module purge
module load bzip2/1.0.8-GCCcore-8.3.0
module load GCC/8.3.0
module load XZ/5.2.4-GCCcore-8.3.0

  5. Go to the amber directory and execute the compilation script

    Code Block
    cd $AMBERHOME
./configure -noX11 gnu

  6. Install and test

    Code Block
    source ./amber.sh
make install
make test

  7. After the installation is successful, you can compile parallel (MPI) versions of Amber20 by load OpenMPI module

    Code Block
    module load OpenMPI/3.1.4-GCC-8.3.0

  8. Then, recompilation with MPI

    Code Block
    ./configure -mpi -noX11 gnu
make install

Info

To compile Amber20 for GPU

  1. Go to the AMBERHOME home directory environment variable

    Code Block
    cd /tarafs/data/project/projxxxx  #change projxxxx to your project ID  
cd amber20_src
export AMBERHOME=$PWD

  2. Load modules and prepare for AMBER20 installation

    Code Block
    module purge
module load bzip2/1.0.8-GCCcore-8.3.0
module load XZ/5.2.4-GCCcore-8.3.0
module load gcccuda/2019b
module load CUDA/10.1.243

  3. Go to the amber directory and execute the compilation script for the installation with cuda

    Code Block
    cd $AMBERHOME
./configure -noX11 -cuda gnu
make install

  4. Testing

    Code Block
    make test.cuda_serial

  5. After the installation is successful, you can compile parallel (MPI) versions with cuda

    Code Block
    module load OpenMPI/3.1.4-gcccuda-2019b
module load NCCL/2.4.8-gcccuda-2019b
export NCCL_HOME=/tarafs/utils/modules/software/NCCL/2.4.8-gcccuda-2019b/

  6. Then, recompilation with MPI for GPU

    Code Block
    ./configure -noX11 -cuda -mpi -nccl gnu
make install