Serial

For when only one CPU is required, generally as part of an job array.

#!/bin/bash -e

#SBATCH --job-name      Delft3D
#SBATCH --time          00:05:00       # Walltime
#SBATCH --mem           512M           # Total Memory
#SBATCH --hint          nomultithread  # Hyperthreading disabled

module load Delft3D

d_hydro test_input.xml

Multi-threaded 

For domain based decompositions.

Use cpus-per-task to allocate resources.

Each subdomain runs in a separate
thread, inside one executable. Limited to one node.

#!/bin/bash -e

#SBATCH --job-name      Delft3D 
#SBATCH --time          00:05:00       # Walltime
#SBATCH --cpus-per-task 4              
#SBATCH --mem           2G             # Total Memory
#SBATCH --hint          nomultithread  # Hyperthreading disabled

module load Delft3D

d_hyrdo test_input.xml

MPI

Domain is split automatically in stripwise partitions. Can run across multiple nodes.

Use ntasks to allocate resources.

Cannot be used in conjunction with:

• DomainDecomposition
• Fluid mud
• Coup online
• Drogues and moving observation points
• Culverts
• Power stations with inlet and outlet in different partitions
• Non-hydrostatic solvers
• Walking discharges
• 2D skewed weirs
• max(mmax,nmax)/npart ≤ 4
• Roller model
• Mormerge
• Mass balance polygons

#!/bin/bash -e

#SBATCH --job-name      Delft3D 
#SBATCH --time          00:05:00       # Walltime
#SBATCH --ntasks        4              
#SBATCH --mem-per-cpu   1G             
#SBATCH --hint          nomultithread  # Hyperthreading disabled

module load Delft3D

srun d_hyrdo test_input.xml