Submit the Large Scale DSO Job: Examples

This section includes examples of submitting a Large Scale DSO job using RSM and a Scheduler.

Note: Functionality featured in the example(s) in this section applies to multiple design types.

Using the RSM Environment

  1. Before submitting the job make sure that the input project is not open in any Desktop window.
  2. For this example, suppose that there are two quad-core machines on your cluster with the names "m1" and "m2". Further suppose that there are two engines per machine, for a total of four parallel engines. Let the number of processors allocated to each engine be 1.
  3. From a command prompt, issue the following command:

    <installation-directory>\v<version>\<platform>\desktopjob.exe -cmd dso -machinelist "list=m1:2,m2:2" -batchoptions \\sjo7na\mxwlprojs\mxwloptions.txt -batchsolve "relay_nominal"Optimetrics:ParametricSetup1" \\sjo7na\mxwlprojs\relay_nominal.aedt

    where the file \\sjo7na\mxwlproj\mxwloptions.txt has the following contents:

    $begin Config

    'Desktop/Settings/ProjectOptions/NumberOfProcessors'=1

    'Maxwell3D/Preferences/SaveBeforeSolving'=0

    'Maxwell/Preferences/HPCLicenceType'='pack'

    #end 'Config'

  4. Assume the above job is assigned ID "jobID".

Using a Scheduler Environment (such as LSF)

  1. Suppose you want to solve variations using four parallel engines, each engine being assigned a single core.
  2. From a command prompt, run the following command:

    bsub -n 4 <installation-directory>\v<version>\<platform>\desktopjob.exe -cmd dso -batchoptions \\sjo7na\mxwlprojs\mxwloptions.txt -batchsolve "relay_nominal:Optimetrics:ParametricSetup1"\\sjo7na\mxwlprojs\relay_nominal.aedt

    where the file \\sjo7na\mxwlproj\mxwlptions.txt has the same contents as the RSM example above.

  3. Assume the above job is assigned an ID "jobid"

For example with 2 nodes of 44 cores each (with multiple distributed variations per machine):

MANUAL: bsub -n 88 <installation-directory>\v<version>\<platform>\desktopjob -cmd dso -machinelist num=8 -monitor -useelectronicsppe=1 -ng -batchoptions " 'Maxwell 3D/NumCoresPerDistributedTask'=11 'Maxwell 3D/RAMLimitPercent'=90 'HPCLicenseType'='Pack'" -batchsolve TeeModel:Optimetrics:ParametricSetup1 OptimTee-v231-1.aedt

AUTO: bsub -n 88 <installation-directory>\v<version>\<platform>\desktopjob -cmd dso -machinelist numcores=88 -auto -NumDistributedVariations 8 -monitor -useelectronicsppe=1 -ng -batchoptions " 'Maxwell 3D/RAMLimitPercent'=90 'HPCLicenseType'='Pack'" -batchsolve TeeModel:Optimetrics:ParametricSetup1 OptimTee-v231-1.aedt

For example, with 16 nodes of 44 cores each (note each distributed variation will span 2 machines):

AUTO: bsub -n 704 <installation-directory>\v<version>\<platform>\desktopjob -cmd dso -machinelist numcores=704 -auto -NumDistributedVariations 8 -monitor -useelectronicsppe=1 -ng -batchoptions " 'Maxwell 3D/RAMLimitPercent'=90 'HPCLicenseType'='Pack'" -batchsolve TeeModel:Optimetrics:ParametricSetup1 OptimTee-v231-1.aedt

Related Topics 

Large Scale DSO for Parametric Analysis

Large Scale DSO Example: Postprocess the Results