ML21230A358

From kanterella
Jump to navigation Jump to search
Lessons Learned - Xlpr Run Optimization
ML21230A358
Person / Time
Issue date: 08/18/2021
From:
NRC/RES/DE
To:
Homiack M
Shared Package
ML21230A354 List:
References
Download: ML21230A358 (11)


Text

Lessons Learned - Run Optimization xLPR User Group Meeting August 18, 2021 1

  • Two xLPR projects at SI thus far:
  • 24 cases, 100k realizations each case
  • ~16 cases, 800k realizations each case
  • 400k per run
  • Efficiency is key to meeting deadlines and maintaining budget!

2 xLPR Use at Structural Integrity Associates (SI)

3 Run Optimization (Brief Recap)

  • Primary goals:
  • Reduce run time (more runs in a given timeframe)
  • Reduce file size (more realizations per run)
  • Reduce overhead in making xLPR runs (reduce time between making runs)

for videos/presentations

Output Settings (1/2)

  • Only allow certain outputs to be saved
  • Reduce frequency of saved outputs 4

Output Settings (2/2) 5

Distributed Processing

  • Allows realizations to be run in parallel across different slave processes
  • GoldSim Pro allows 4 slaves, Distributed Processing has no limit (but should only use # of cores - 1)
  • Nearly linear improvement in runtimes (twice the slaves means half the run time)
  • GSP Slaves can be opened one at a time in the Run utility 6

Desktop Shortcuts and Batch Files (1/3)

  • Slaves can also be run as a desktop shortcut 7

Desktop Shortcuts and Batch Files (2/3) 8 Target: C:\\Program Files (x86)\\GTG\\GoldSim 11.1\\GoldSim.exe -s Start in: C:\\Program Files (x86)\\GTG\\GoldSim 11.1

Desktop Shortcuts and Batch Files (3/3)

  • A batch file can be made to open multiple slave processes instantaneously
  • Replace n with number of slaves to open
  • Suggest closing and reopening all slaves between larger runs 9

Using Distributed Processing Properly

  • Must disable run in separate process
  • Class View > External > Double Click each DLL 10
  • Allows each slave to operate independently

Demonstration 11