AllocCPUS

AllocCPUS is a Slurm job environment variable that indicates the number of CPUs allocated to the current job on the node where the job’s processes are running. It is set by the Slurm controller at job start and is exposed to tasks launched by srun or sbatch so that applications can adapt to the available parallelism.

The value represents the per-node CPU allocation for the job. In typical configurations, each node assigned to a multi-node job will have AllocCPUS equal to the number of CPUs reserved on that node for the job. For example, if a user requests 16 CPUs per node on a two-node job, each node’s AllocCPUS would usually be 16. The variable is useful for configuring multi-threaded applications to match the actual resources granted.

AllocCPUS can be accessed from within job scripts or applications as an environment variable. In a shell script, one might use it to set the number of threads, such as exporting OMP_NUM_THREADS=$AllocCPUS for OpenMP workloads. It is often used in conjunction with other Slurm-provided variables that describe the overall job resource layout, though exact names and availability can vary by Slurm version and configuration.

Notes and limitations: not all environments guarantee the AllocCPUS variable, and some workflows rely on alternative Slurm variables (for example SLURM_JOB_CPUS or SLURM_CPUS_ON_NODE) to derive parallelism. When in doubt, consult the local Slurm documentation or print the environment in a test job to confirm which variables are exported.

See also Slurm, environment variables, OpenMP, parallel computing.