Tech Articles

SMP with SOL 103


Tech Articles ID    KB8022077
Status:    Published
Published date:    04/23/2015
Updated:    06/29/2015
Reported In:   MSC Nastran - MSC Nastran - Standard (SOL 1 through 200)
 

Abstract (Question)

Summary

  • SMP performance improvement is highly model dependent. Understanding expectations on your job is important.
  • SMP works with DMP. As a general rule, set SMP to the number of cores per socket and DMP to the number of sockets.

Description (Answer)

Details

Below is a chart for various models with varying number of SMP processors. Performance improvement varies.

In general, the work done in Parallel for SOL 103 is Decomposition, Forward Backward Substitution and Multiply/ADD.

Job N dof N 2-D elements N 3-D elementes N Roots Method
xx1bmd0 8,627,970 31,838 1,526,072 536 ACMS
xx0wmd0 3,799,278 704,415 211,596 1221 ACMS
x351 10,956,143 150,3062 742,660 50 Lanczos
xx0vmd0 2,604,102 0 561,545 50 Lanczos

 

Commands

SMP parallel is activated with either smp=n on the command line or with the following in the nastran section of your input file:

     NASTRAN PARALLEL=n

Where "n" is the number of CPUs

Caveats

 

  • Some analysis may be slower with SMP
  • Using cores from mulitple sockets ( more than exist on a single socket ) may run slower with SMP

Rate this Page
Rate the quality of this article from 5 (high) to 1 (low):
1
2
3
4
5
 
How can we improve this?