CUDA Program Using Singularity Environment

Step 1: Create Singularity Image (CUDA)

To create SIngularity image (CUDA), run below command in the SSH terminal (CLI), for tutorial in accessing CLI, please refer to SSH Shell Access to EdUHK HPC Platform and Cluster (Web-based Shell Access)

# Create new directory (if needed)
$ mkdir /home/${USER}/singularity_image

# Pull docker image by singularity
$ cd /home/${USER}/singularity_image
$ module load singularity
$ singularity pull docker://nvidia/cuda:12.8.0-devel-ubuntu24.04

# List pull image
# (Example:cuda_12.8.0-devel-ubuntu24.04.sif)
$ ls –al

Step 2: Prepare CUDA Program Source Code

For the example source code please refer to CUDA Program Source Code (Example)

Step 3: Prepare job template script

Pre-configured template script path → /home/$USER/job_template/slurm_job/cuda_singularity.sh

#!/bin/bash
#SBATCH --job-name=cuda_singularity ## Job Name
#SBATCH --partition=shared_gpu_l40 ## Partition for Running Job
#SBATCH --nodes=1 ## Number of Compute Node
#SBATCH --ntasks=1 # Number of Tasks
#SBATCH --cpus-per-task=2 ## Number of CPU per task
#SBATCH --time=60:00 ## Job Time Limit (i.e. 60 Minutes)
#SBATCH --gres=gpu:l40:1 # Number of GPUs per node (i.e. 1 x l40 GPU)
#SBATCH --mem=10GB ## Total Memory for Job
#SBATCH --output=./%x%j.out ## Output File Path
#SBATCH --error=./%x%j.err ## Error Log Path

## Initiate Environment Module
source /usr/share/modules/init/profile.sh

## Reset the Environment Module components
module purge

## Load Module
module load singularity

## Run user command
singularity run --nv \
--bind /home/${USER}/job_template/C \
/home/${USER}/singularity_image/cuda_12.8.0-devel-ubuntu24.04.sif \
nvcc -o pi_cuda /home/${USER}/job_template/C/cuda_pi.cu && \
./pi_cuda

## Clean up
rm pi_cuda

## Clear Environment Module components
module purge

Step 4: Submit HPC Job

Guides for submitting HPC job, please refer to: HPC Job Submission (For CLI) and HPC Job Submission (For Web Portal)

Step 5: Remove Singularity Image (Optional)

$ rm /home/${USER}/singularity_image/<image file>.sif