Following Cadence Design Systems’ acquisition of the Design & Engineering (D&E) business, the D&E Community Forum on the Nexus Community Center has been migrated in the Cadence Physical System Simulation Community. For more details and if you have any questions, please refer to the FAQ.

Stats

  • State Not Answered
  • Replies 5
  • Subscribers 3
  • Views 27
  • Members are here 0
More Content

Is there a way to speed up or control batch solver file generation?

SHonodel
I frequently generate hundreds or thousands of adm (and associated) files via mdi insight build command for a Monte Carlo analysis.  In this current​ MC analysis, my model is pretty large and has 3 flex bodies and 2MB of xmt_txt geometry so it takes about 15 seconds to generate each case.  So it takes 3 hours to generate 750 cases.  However, I am not changing the geometry between these cases so I assume the xmt_txt files are identical, right?  Is there an option to turn off the generation of the xmt_txt file and tell Solver (via .acf file, I assume) what xmt_txt file to use?
  • 0
    First things that'd come to my mind are not all ADAMS related:
    • How about speeding up file IO, i.e. installing a SSD ?
    • Are you solving the adm's parallel or sequential ? In the latter case: Have you tried internal solver instead of external ? That'd make the adm/xmt exports obsolete.
    • Is there any geometry used in calculation (like contacts)? If not, you could delete/deactivate the geometry before starting the batch runs which will avoid exporting any xmt's.
     
    And then a homework for MSC's development:
    If I defined geometry/parts a template/model by importing one ore more parasolid file(s), why "convert" that to a single parasolid on exporting the adm ?
    How about just keeping the original file references in the adm and save the re-write of huge geometry files ? I agree that in some cases I deliberately wanted this "conversion" and thus I'd vote for a setting to choose between the old and the newly proposed behaviour.
     
     
     
  • 0
    Hi Guys,
     
    There is an option in Adams/View for multi-run studies named "One Parasolid" that I think might help. Find it:
    Settings -> Solver -> Output
    then 'Output Category' is 'Database Storage' and you'll see the "One Parasolid" option that applies to multi-run studies.
     
    When this is off (default) you get an xmt file for every adm file. When this is on you should only get 1 xmt file. I just tested it on a *simple* model and it does what you want. I am hopeful that it also works on your complex model. Obviously we do some logic to determine if it's ok to do this; I don't know the extent of that logic so I'm hoping my simple test represents what happens in your model..
     
    HTH,
    Kent
     
  • 0
    Kent, thanks for that info - it'll turn out handy in some of my own batch analysis if it works as advertised.
  • 0
    In addition to the above fixes, you can speed this up significantly using brute force by modifying the "mdi insight build" macro to accept a "startCase" and "endCase" parameter and build a subset of your workspace. You can then run multiple parallel aview jobs that export subsets of your Monte Carlo (1-100 in job 1, 101 - 200 in job 2 etc). You have to write some sort of postprocessor that re-assembles the acf/adm files after all the parallel build jobs are complete, but running the build in parallel speeds the build process up significantly.
     
    No matter what you do the job will write out .mtx file for every flexbody in every case, which is extremely wasteful. I hope they can add a similar flag to have a "One MTX" option in the future.
  • 0
    Thanks, everybody, for weighing in.  Kent, thanks for the great news!  If only I had looked around a bit when I was in Database Storage window I could have saved myself hours of case generation and data transfer (slower via VPN).​ 
     
    Martin - SSD would be nice but we're no longer allowed to attach drives/etc. to company computers.  We went to batch Solver runs specifically to speed up solution time, which is much longer than case generation/transfer time.  And, yes, my model has lots of contacts so xmt_txt file is required.
     
    Michael - That's a good idea.  I might resort to that if I don't get enough speed-up from One Parasolid option.
Cadence Guidelines

Community Guidelines

The Cadence Design Communities support Cadence users and technologists interacting to exchange ideas, news, technical information, and best practices to solve problems and get the most from Cadence technology. The community is open to everyone, and to provide the most value, we require participants to follow our Community Guidelines that facilitate a quality exchange of ideas and information. By accessing, contributing, using or downloading any materials from the site, you agree to be bound by the full Community Guidelines.