Dear all,
The amazing mps code (mps_evolve and mps_optim) works very well with OpenMP on a node of a cluster when I start the task by hand. It can spawn many threads and calculate very fast. But when I wrote a sbatch script to commit a job to cluster by slurm, the mps applications can only use one thread and calculate very slowly. The sbatch script is shown as
-----------------------------------------------------------------
#!/bin/bash
#SBATCH --job-name=test_alps_omp
#SBATCH --output=test_alps_omp.out
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=4
export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK
mps_evolve test_alps_omp.in.xml
-----------------------------------------------------------------
Does anyone have any idea about it? Thanks a lot!
Beat regards,
Rongyang Sun
-----------------------------------------------------------------
Rongyang Sun
Ph.D. Candidate
Theoretical Quantum Matter Physics
Institute for Advanced Study, Tsinghua University, Beijing 100084, China
E-mail: sunry15(a)mails.tsinghua.edu.cn; sun-rongyang(a)outlook.com