3.4. Submitting jobs to Shared-Queue¶

This describes how to submit jobs to the shared queue. The shared queue is a queue used to share nodes with other jobs.

For interactive work, please use IP_001 or IA_001. If you need to refer to GPU information, please use IA_001.

When executing small jobs, please use CP_001 or CA_001. If the job requires a GPU, please use CA_001.

When performing debugging tasks, please use DP_002 or DA_002. If the job requires a GPU, please use DA_002.

Jobs executed in CA_001/DA_002 are allocated 1 CPU and 1 GPU by default.

Jobs executed outside of CA_001/DA_002 are allocated 1 CPU by default.

The allocation of CPU and GPU can be specified through the options of the qsub command.

Please see List of queues for the maximum limit that can be specified.

3.4.1. Usage¶

・CA_001/DA_002

qsub -q queue name [ -I ] [ -l select=1[:ncpus=number of CPUs][:ngpus=number of GPUs] [program]

・Outside of CA_001/DA_002

qsub -q queue name [ -I ] [ -l select=1[:ncpus=number of CPUs] [program]

・Example queue:IP_001, CPU:2, interactive mode

user01@super1:~> qsub -I -q IP_001 -l select=1:ncpus=2
qsub: waiting for job 10.cmspbs1 to start
qsub: job 10.cmspbs1 ready

・Example queue:CP_001, CPU:28, program:a.out

#!/bin/sh
#PBS -l select=1:ncpus=28
#PBS -q CP_001
#PBS -N sample

mpirun -np 28 -ppn 28 ./a.out > result.out 2> result.err