Intel® Moderncode for Parallel Architectures
Support for developing parallel programming applications on Intel® Architecture.

A question about the performance of IntelMPI

ye__xiang
Beginner
847 Views

Dear ALL, 

      We have an IntelMPI problem and we need help. Recently, we used the latest Parallel Studio(parallel_studio_xe_2018_update2_cluster_edition) to compile an simulation package(VASP). When we run the program, it is found that the load rate (%CPU) of some cores change repeatedly from 20% to 90%, and the status of some cores changes from "R" to "D".(you can referred to the attachment, ) Is it reasonable? If not, how can we fix it.Thanks!

The platform detail: each node have 2 cpus (Intel(R) Xeon(R) Gold 6132 CPU @ 2.60GHz), nodes are connected by infiniband. The pbs file used is shown below

#! /bin/bash 
#PBS -N test1_yx
#PBS -l nodes=2:ppn=16
#PBS -l walltime=240:00:00
#PBS -j oe
#PBS -q blade
n_proc=$(cat $PBS_NODEFILE | wc -l)

cd $PBS_O_WORKDIR

mpirun -np ${n_proc}   -hostfile $PBS_NODEFILE  /public/sourcecode/vasp/vasp.5.4.4/bin/vasp_std_constr



Bests!
Xiang YE

0 Kudos
1 Reply
TimP
Honored Contributor III
847 Views

I don't see why this was bumped up to current status.

This looks highly dependent on the application and your disk system.  Making your D access gain from parallelism is a separate and usually more difficult problem, but it can't be characterized as an MPI problem, let alone an Intel MPI problem.  If you did have a problem with Intel MPI, the companion forum on cluster computing would be the place to ask.

0 Kudos
Reply