Job Builder
Hardware Request
The number of tasks are the number of separate processes you plan to launch in parallel. For MPI jobs, this is number of MPI processes (the argument passed with '-np'). If unsure, leave at 1. Note that increasing this number will not guarantee a faster runtime.
The number of CPU cores per task. If unsure, leave at 1. Note that increasing this number will not guarantee a faster runtime.
This specifies the amount of memory you need per task. The different units are kilobytes(kb), megabytes(mb), gigabytes(gb), and terabytes(tb). If using multiple tasks, this is automatically converted to memory per cpu.
Feature Constraints
Enables high-speed communication between multiple GPUs. Only select if your programs support using multiple GPUs in parallel.
Users can specify additional constraints required for their job. Please visit the complete list of feature constraints for more information. Leave blank if none are needed. List each constraint and separate them with a comma. Exclude spaces.
Ex: gpu_interconnect_pcie,chip_manufacturer_intel
Job Details
Walltime is the maximum amount of real-world time your job should take to complete. Once the time expires, your job will be terminated if it is still running.
Users with multiple slurm accounts can select which one this job is assigned to. This account will also handle the billing of the job. For a complete list of your accounts, use the command sacctmgr show assoc user=USERNAME format=account%64,user
in a Palmetto 2 terminal. Leave blank for default account.
Users can give their jobs a unique name to make it easier to find in reports. The specified name will appear along with the job id number when querying running jobs on the system. The default name is the name of the "command" specified on the command line.
#!/bin/bash
# Submit this script with: sbatch FILENAME
#SBATCH --ntasks 1 # number of tasks
#SBATCH --cpus-per-task 1 # number of cpu cores per task
#SBATCH --time 1:00:00 # walltime
#SBATCH --mem 1gb # amount of memory per CPU core (Memory per Task / Cores per Task)
#SBATCH --nodes 1 # number of nodes
# LOAD MODULES, INSERT CODE, AND RUN YOUR PROGRAMS HERE