Title: How should I request memory for a SLURM job?

Goal:
I need to run a Python application that uses approximately 40 GB of memory.

What I have tried:
I submitted the job using the following option:

#SBATCH --mem=40G

Question:
Does --mem=40G request 40 GB for the entire job or 40 GB per CPU?

Documentation consulted:
[Add a link to the relevant cluster documentation]