View in browser

K2realm, LLC multicore tolerancing

Feedback / Suggest Topics

Multi-core is the default

  • Both Monte Carlo and finite difference tolerancing automatically use available cores
  • Run-times have been reduced by 10X using a 16-core PC (~$5k for modern desktops)
  • Users report a 2-hour run was reduced to 10 minutes, and a 20 hour run is now 2 hours!
  • Some users have previously limited runs to 500 trials due to time, but can now run 1,000’s of trials in less time for more accuracy
  • No extra setup is required, and it works with current performance and compensation sequence files, as well as user-defined
  • Combined data results are saved to a file for further processing if desired
K2realm, multicore tolerancing with CODE V


Easily Add Multiple Custom Metrics

  • Easily evaluate multiple metrics (no need for ^imagqual index manipulation)
  • Customize the view to see each metric side-by-side, individually, or plotted on the same chart

Immediately Visualize Tolerance Drivers

  • Visualize tolerance drivers per surface via bar charts (TOR and finite difference)
  • Sort all tolerances to quickly see worst drivers
  • Eliminates need to search through text output, or copy/paste to Excel

Use Multi-core with Custom Scripts

  • Any long running script with independent loops (e.g. over field, zoom, or list of lenses) can take advantage of multiple cores
  • Special variables pass info to the custom macro on how many cores to use, and the current core #
  • The macro loops are modified to save data based on this info
  • We can help make these modifications on existing macros
  • The Extensions handle the distribution of the work, and collect all data when finished

Incorporate MATLAB into Analyses

  • Techniques have been developed to incorporate MATLAB in custom scripts
  • The basic procedure is the following,

    • Call CODE V options to generate data
    • Send this data to MATLAB
    • Run a custom MATLAB script to process the data
    • Send the MATLAB results back to CODE V
  • These steps are executed with each trial or loop iteration
Schedule a Demo for More Info