====== Amber ======
/* Description */
//Assisted Model Building with Energy Refinement// : "Amber" refers to two things: a set of molecular mechanical force fields for the simulation of biomolecules (AmberTools, software in the  public domain) ; and a package of molecular simulation programs.

  * Site web : http://ambermd.org/
  * Versions installées :
    * AmberTools (libre) : v14
    * Amber (payant) : v12 avec AmberTools v13

La compilation a été faite avec le support OpenMP (utilisable pour certains programmes comme ''paramfit'' ou ''NAB'') et MPI. Les versions MPI des programmes sont postfixés avec ''.MPI'', par exemple ''sandr.MPI'' : il s'agit de l'installation standard d'Amber, la documentation d'Amber s'applique.

===== Citation =====
Si vous utilisez le support GPU : vous devez citer des articles dans vos publications. Voir http://ambermd.org/gpus12/ section //Citing the GPU Code//
===== Utilisation =====
==== Sélection de la version ====
Pour sélectionner la version voulue : utiliser les [[..:modules]]

Pour utiliser Amber 12 avec AmberTools 13 :
  module load amber/12

Pour utiliser Amber 12 avec AmberTools 14 :
  module load amber/12-tools14

==== Travailler avec slurm ====
/* Faut-il utiliser des fichiers de commandes spéciaux ? */
/* Exemples de batch spécifiques */

Si vous utilisez un des programmes Amber parallélisé via OpenMP, vous devez suivre l'[[:slurm:slurm-multithread | exemple Slurm pour OpenMP]] pour en tirer profit.

==== Version GPU ====

Une version GPU est compilée (http://ambermd.org/gpus12/). L'exécutable s'appelle ''pmemd.cuda'' et doit être exécuté sur le neoud GPU, voir les instructions s'appliquant à CALI sur [[:gpu:utilisation | la page GPU]].

Vous devez charger en plus le module CUDA version 5.0 : 
  module load nvidia/cuda/5.0

===== Exemples =====

==== Serial ====


  #!/bin/bash
  #SBATCH --nodes=1
  #SBATCH --ntasks=1
  #SBATCH --time=02:00:00
  #SBATCH --partition=normal
  #SBATCH --cpus-per-task=1
  #SBATCH --mem-per-cpu=2000 
  topname=del
  crdname=del
  CURR=$(pwd)
  RESTART=YES
  STEP=0
  FINAL=1
  module load amber/16-patched-10062016
  # MD are performed in Scratch 
  # Export running directectory
  export RUN_DIR="${HOME}/scratch/Amber/run.${SLURM_JOB_ID}.amber16.${topname}"
  # Link the scratch directory to the current directory 
  # Link name is related to JOB ID and NOT TO JOBNAME --> Prevent issues related to new MDs with same name
  ln -sfn ${RUN_DIR} ${SLURM_JOB_ID}.results
  echo ${RUN_DIR} 
  # Create Scratch directory
  mkdir -p ${RUN_DIR}
  # Copy Topology and CRD names into Scratch directory
  cp ${topname}.prmtop *.in ${RUN_DIR}
  # Copy starting CRD files according to restart procedure or not
  if [ $RESTART = 'YES' ]; then
      cp ${topname}_md${STEP}.rst ${RUN_DIR}/
  else
      STEP=0
      cp ${topname}.inpcrd ${RUN_DIR}/
  fi
  # Enter into Scratch directory
  cd ${RUN_DIR}
  # Run calculation 
  # Keep in mind: Never run a long MD into a single trajectory file - Split it into several trajectories
  for i in `seq $STEP 1 $FINAL`
  do
      j=$(( $i + 1 ))
      # Run MD
      pmemd -O -i md.in -o ${crdname}_md${j}.out -p $topname.prmtop -c ${crdname}_md${i}.rst -r ${crdname}_md${j}.rst -x ${crdname}_${j}.mdcrd
      # Compress mdcrd file to prevent storage issues and remove the uncompressed trajectory
      tar -zcvf ${crdname}_${j}.mdcrd.tar.gz ${crdname}_${j}.mdcrd
      rm -rf ${crdname}_${j}.mdcrd
  done

==== MPI ====

  #!/bin/bash
  #SBATCH --nodes=1-X
  #SBATCH --ntasks=X
  #SBATCH --time=XX:00:00
  #SBATCH --partition=normal
  #SBATCH --cpus-per-task=1
  #SBATCH --mem-per-cpu=XXX 
  topname=del
  crdname=del
  CURR=$(pwd)
  RESTART=YES
  STEP=0
  FINAL=1
  module load amber/16-patched-10062016
  # MD are performed in Scratch 
  # Export running directectory
  export RUN_DIR="${HOME}/scratch/Amber/run.${SLURM_JOB_ID}.amber16.${topname}"
  # Link the scratch directory to the current directory 
  # Link name is related to JOB ID and NOT TO JOBNAME --> Prevent issues related to new MDs with same name
  ln -sfn ${RUN_DIR} ${SLURM_JOB_ID}.results
  echo ${RUN_DIR} 
  # Create Scratch directory
  mkdir -p ${RUN_DIR}
  # Copy Topology and CRD names into Scratch directory
  cp ${topname}.prmtop *.in ${RUN_DIR}
  # Copy starting CRD files according to restart procedure or not
  if [ $RESTART = 'YES' ]; then
      cp ${topname}_md${STEP}.rst ${RUN_DIR}/
  else
      STEP=0
      cp ${topname}.inpcrd ${RUN_DIR}/
  fi
  # Enter into Scratch directory
  cd ${RUN_DIR}
  # Run calculation 
  # Keep in mind: Never run a long MD into a single trajectory file - Split it into several trajectories
  for i in `seq $STEP 1 $FINAL`
  do
      j=$(( $i + 1 ))
      # Run MD
      srun premed.MPI -O -i md.in -o ${crdname}_md${j}.out -p $topname.prmtop -c ${crdname}_md${i}.rst -r ${crdname}_md${j}.rst -x ${crdname}_${j}.mdcrd
      # Compress mdcrd file to prevent storage issues and remove the uncompressed trajectory
      tar -zcvf ${crdname}_${j}.mdcrd.tar.gz ${crdname}_${j}.mdcrd
      rm -rf ${crdname}_${j}.mdcrd
  done

==== GPU ====

  #!/bin/bash
  #SBATCH --nodes=1
  #SBATCH --time=02:00:00
  #SBATCH --partition=gpu
  #SBATCH --cpus-per-task=1
  #SBATCH --gres=gpu
  topname=del
  crdname=del
  CURR=$(pwd)
  RESTART=YES
  STEP=0
  FINAL=1
  module load amber/16-patched-10062016
  module load nvidia/cuda/7.5
  # MD are performed in Scratch 
  # Export running directectory
  export RUN_DIR="${HOME}/scratch/Amber/run.${SLURM_JOB_ID}.amber16.${topname}"
  # Link the scratch directory to the current directory 
  # Link name is related to JOB ID and NOT TO JOBNAME --> Prevent issues related to new MDs with same name
  ln -sfn ${RUN_DIR} ${SLURM_JOB_ID}.results
  echo ${RUN_DIR} 
  # Create Scratch directory
  mkdir -p ${RUN_DIR}
  # Copy Topology and CRD names into Scratch directory
  cp ${topname}.prmtop *.in ${RUN_DIR}
  # Copy starting CRD files according to restart procedure or not
  if [ $RESTART = 'YES' ]; then
      cp ${topname}_md${STEP}.rst ${RUN_DIR}/
  else
      STEP=0
      cp ${topname}.inpcrd ${RUN_DIR}/
  fi
  # Enter into Scratch directory
  cd ${RUN_DIR}
  # Run calculation 
  # Keep in mind: Never run a long MD into a single trajectory file - Split it into several trajectories
  for i in `seq $STEP 1 $FINAL`
  do
      j=$(( $i + 1 ))
      # Run MD
      pmemd.cuda -O -i md.in -o ${crdname}_md${j}.out -p $topname.prmtop -c ${crdname}_md${i}.rst -r ${crdname}_md${j}.rst -x ${crdname}_${j}.mdcrd
      # Compress mdcrd file to prevent storage issues and remove the uncompressed trajectory
      tar -zcvf ${crdname}_${j}.mdcrd.tar.gz ${crdname}_${j}.mdcrd
      rm -rf ${crdname}_${j}.mdcrd
  done