Using GPUs with Slurm

For general advice on job scheduling, see Running jobs.

Available hardware

These are the node types containing GPUs currently available on Cedar and Graham:

# of Nodes	Node type	CPU cores	CPU memory	# of GPUs	GPU type	PCIe bus topology
114	Cedar Base GPU	24	128GB	4	NVIDIA P100-PCIE-12GB	Two GPUs per CPU socket
32	Cedar Large GPU	24	256GB	4	NVIDIA P100-PCIE-16GB	All GPUs associated with the same CPU socket
160	Graham Base GPU	32	128GB	2	NVIDIA P100-PCIE-12GB	One GPU per CPU socket

Single-core job

If you need only a single CPU core and one GPU:

File : gpu_serial_job.sh

#!/bin/bash
#SBATCH --account=def-someuser
#SBATCH --gres=gpu:1              # Number of GPUs (per node)
#SBATCH --mem=4000M               # memory (per node)
#SBATCH --time=0-03:00            # time (DD-HH:MM)
./program                         # you can use 'nvidia-smi' for a test

Multi-threaded job

For GPU jobs asking for multiple CPUs in a single node:

File : gpu_threaded_job.sh

#!/bin/bash
#SBATCH --account=def-someuser
#SBATCH --gres=gpu:1              # Number of GPU(s) per node
#SBATCH --cpus-per-task=6         # CPU cores/threads
#SBATCH --mem=4000M               # memory per node
#SBATCH --time=0-03:00            # time (DD-HH:MM)
export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK
./program

On Cedar, we recommend that multi-threaded jobs use no more than 6 CPU cores for each GPU requested. On Graham, we recommend no more than 16 CPU cores for each GPU.

MPI job

File : gpu_mpi_job.sh

#!/bin/bash
#SBATCH --account=def-someuser
#SBATCH --gres=gpu:4              # Number of GPUs per node
#SBATCH --nodes=2                 # Number of nodes
#SBATCH --ntask=48                # Number of MPI process
#SBATCH --cpus-per-task=1         # CPU cores per MPI process
#SBATCH --mem=120G                # memory per node
#SBATCH --time=0-03:00            # time (DD-HH:MM)
export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK
srun ./program

Whole nodes

If your application can efficiently use an entire node and its associated GPUs, you will probably experience shorter wait times if you ask Slurm for a whole node. Use one of the following job scripts as a template.

Scheduling a GPU node at Graham

File : graham_gpu_node_job.sh

#!/bin/bash
#SBATCH --nodes=1
#SBATCH --gres=gpu:2
#SBATCH --ntasks-per-node=32
#SBATCH --mem=127000M
#SBATCH --time=3:00
#SBATCH --account=def-someuser
nvidia-smi

Scheduling a Base GPU node at Cedar

File : cedar_gpu_node_job.sh

#!/bin/bash
#SBATCH --nodes=1
#SBATCH --gres=gpu:4
#SBATCH --exclusive
#SBATCH --mem=125G
#SBATCH --time=3:00
#SBATCH --account=def-someuser
nvidia-smi

Scheduling a Large GPU node at Cedar

There is a special group of large-memory GPU nodes at Cedar which have four Tesla P100 16GB cards each. (Other GPUs in the cluster have 12GB.) The GPUs in a large-memory node all use the same PCI switch, so the inter-GPU communication latency is lower, but bandwidth between CPU and GPU is lower than on the regular GPU nodes. The nodes also have 256 GB RAM instead of 128GB. You may only request these nodes as whole nodes, therefore you must specify --gres=gpu:lgpu:4. Note that the maximum run-time for the large-memory GPU nodes on Cedar used to be 24 hours, this is not the case any more.

File : large_gpu_job.sh

#!/bin/bash
#SBATCH --nodes=1 
#SBATCH --gres=gpu:lgpu:4   
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=24    # There are 24 CPU cores on Cedar GPU nodes
#SBATCH --mem=0               # Request the full memory of the node
#SBATCH --time=3:00
#SBATCH --account=def-someuser
hostname
nvidia-smi

Packing single-GPU jobs within one SLURM job

If you need to run four single-GPU programs or two 2-GPU programs for longer than 24 hours, GNU Parallel is recommended. A simple example is given below:

cat params.input | parallel -j4 'CUDA_VISIBLE_DEVICES=$(({%} - 1)) python {} &> {#}.out'

In this example the GPU ID is calculated by subtracting 1 from the slot ID {%}. {#} is the job ID, starting from 1.

A params.input file should include input parameters in each line like:

code1.py
code2.py
code3.py
code4.py
...

With this method, users can run multiple tasks in one submission. The -j4 parameter means GNU Parallel can run a maximum of four concurrent tasks, launching another as soon as each one ends. CUDA_VISIBLE_DEVICES is used to ensure that two tasks do not try to use the same GPU at the same time.

Message Passing Interface