Poor Run Time Despite using SPMD

Crashphys
Crashphys Altair Community Member
edited October 2020 in Community Q&A

Hi,

 

I am using SPMD on the attached model with:

 

-optskip -mpi i -ldm -np 4

 

Without using LDM, I get a simulation time of 19 seconds. With it, I get 18 seconds. Is there anything else I can do to speed up the simulation? Am I perhaps using it incorrectly? I used the attached guide to start the SPMD run. I also tried SMP and it did not give any noticeable advantages.

Unable to find an attachment - read this blog

Tagged:

Answers

  • Rahul Rajan_21763
    Rahul Rajan_21763 New Altair Community Member
    edited December 2019

    Could you try this in the OptiStruct latest version 2019.0 or latest point release with 2019.2?

  • Crashphys
    Crashphys Altair Community Member
    edited December 2019

    Hi,

     

    I got my sysadmin to update me to OS 2019. It currently runs slower than it did on 2017, please advise what the issue could be. The one labelled with _013 was run on OS2017 2 days ago. Without the label was run this morning with OS2019. I also have attached a run where I used one core and -np 4.

     

    It appears that CPU time is not improved, however wall time increases. Is there a reason for this on my end?

     

    Something which did not occur to me: is SPMD or SMP recommended for a single Linear Static analysis? I have four load cases. Is this something which can be effectively parallelized? 

     

    Thanks

    Unable to find an attachment - read this blog

  • Q.Nguyen-Dai
    Q.Nguyen-Dai Altair Community Member
    edited December 2019

    @Crashphys: Parallel run (SMP or SPMD) is for big analysis. When you use this method for small (18, 19sec in your case) analysis, there's NO sense. Try with a bigger analysis, about 4~5h of CPU.

     

Categories