Difference between revisions of "LS-DYNA"

From HPC
Jump to: navigation , search
m (Chris.collins moved page Applications/LS-DYNA to LS-DYNA: To edit)
(Usage Examples)
Line 33: Line 33:
 
</pre>
 
</pre>
 
Note: Due to licensing restrictions an LS-DYNA application can only run on 1 CPU
 
Note: Due to licensing restrictions an LS-DYNA application can only run on 1 CPU
 +
 +
===Parallel Processing Options===
 +
There are two approaches to using multiple processing cores within a node with LS-DYNA. The first is Shared Memory Parallel processing which uses ncpu=X to define the number of cores to use. The alternative is Massively Parallel Processing (MPP) which uses the flag -np=X. To use MPP, you also need to use the -dis flag to enable distributed processing. The difference between these settings for one sample job can be seen below:
 +
 +
<pre>
 +
Original run: lsdyna memory=${memory}M i=all.k ncpu=5
 +
estimated total cpu time          =    290065 sec (      80 hrs 34 mins)
 +
estimated cpu time to complete    =    290061 sec (      80 hrs 34 mins)
 +
estimated total clock time        =    59061 sec (      16 hrs 24 mins)
 +
estimated clock time to complete  =    59059 sec (      16 hrs 24 mins)
 +
 +
Run using -dis flag: lsdyna memory=${memory}M -dis i=all.k ncpu=5
 +
estimated total cpu time          =    38261 sec (      10 hrs 37 mins)
 +
estimated cpu time to complete    =    38260 sec (      10 hrs 37 mins)
 +
estimated total clock time        =    45861 sec (      12 hrs 44 mins)
 +
estimated clock time to complete  =    45859 sec (      12 hrs 44 mins)
 +
 +
Run using "lsdyna -dis memory=${memory}M pr=aa_r_dy i=all.k -np 5"
 +
estimated total cpu time          =    20336 sec (      5 hrs 38 mins)
 +
estimated cpu time to complete    =    20336 sec (      5 hrs 38 mins)
 +
estimated total clock time        =    22135 sec (      6 hrs  8 mins)
 +
estimated clock time to complete  =    22134 sec (      6 hrs  8 mins)
 +
</pre>
  
 
==Further Information==
 
==Further Information==
 
Forum Support: [http://hpc.phpbb.hull.ac.uk/viewforum.php?f=10&sid=4bdc913307b103754005e4b827bc8a18 Viper Ansys Forum]<br/>
 
Forum Support: [http://hpc.phpbb.hull.ac.uk/viewforum.php?f=10&sid=4bdc913307b103754005e4b827bc8a18 Viper Ansys Forum]<br/>
 
Ansys Website: [http://www.ansys.com Ansys]
 
Ansys Website: [http://www.ansys.com Ansys]

Revision as of 09:41, 30 January 2018

Application Details

  • Description:
  • Versions: V17.0, V17.2
  • Module names: ansys/v170, ansys/v172
  • License:

Usage Examples

Job Submission Script

#!/bin/bash	
#SBATCH -c 1	# Number of CPUS per Node, Must be set to 1
#SBATCH --mem-per-cpu=4G  # Sets memory per CPU to 4GB
#SBATCH -J LS-DYNA  # Name of Job
#SBATCH -p compute # Use compute partition
#SBATCH -o %N.%j.%a.out  # Output file name
#SBATCH -e %N.%j.%a.err   # Error file name

# Remove all loaded modules
module purge                             

# Load Ansys V17.2 & Intel MPI
module add ansys/v172            
module add intel/mpi/64/5.1.3.181  

# Divide memory between cores
memory=$(($SLURM_MEM_PER_CPU/8)) 

# This is the run command Note i specifies the input file, memory the memory available and ncpu the number of CPUs used
lsdyna i=all.k memory=${memory}M ncpu=$SLURM_CPUS_PER_TASK 
[username@login01 lsdyna5]$ sbatch ls-dyna.job
Submitted batch job 306050

Note: Due to licensing restrictions an LS-DYNA application can only run on 1 CPU

Parallel Processing Options

There are two approaches to using multiple processing cores within a node with LS-DYNA. The first is Shared Memory Parallel processing which uses ncpu=X to define the number of cores to use. The alternative is Massively Parallel Processing (MPP) which uses the flag -np=X. To use MPP, you also need to use the -dis flag to enable distributed processing. The difference between these settings for one sample job can be seen below:

Original run: lsdyna memory=${memory}M i=all.k ncpu=5
 estimated total cpu time          =    290065 sec (      80 hrs 34 mins)
 estimated cpu time to complete    =    290061 sec (      80 hrs 34 mins)
 estimated total clock time        =     59061 sec (      16 hrs 24 mins)
 estimated clock time to complete  =     59059 sec (      16 hrs 24 mins)

Run using -dis flag: lsdyna memory=${memory}M -dis i=all.k ncpu=5
 estimated total cpu time          =     38261 sec (      10 hrs 37 mins)
 estimated cpu time to complete    =     38260 sec (      10 hrs 37 mins)
 estimated total clock time        =     45861 sec (      12 hrs 44 mins)
 estimated clock time to complete  =     45859 sec (      12 hrs 44 mins)

Run using "lsdyna -dis memory=${memory}M pr=aa_r_dy i=all.k -np 5"
 estimated total cpu time          =     20336 sec (       5 hrs 38 mins)
 estimated cpu time to complete    =     20336 sec (       5 hrs 38 mins)
 estimated total clock time        =     22135 sec (       6 hrs  8 mins)
 estimated clock time to complete  =     22134 sec (       6 hrs  8 mins)

Further Information

Forum Support: Viper Ansys Forum
Ansys Website: Ansys