Large Scale DSO Command Line Syntax

Large Scale DSO operates through a non-graphical batch application called desktopjob. You can run the desktopjob command line to perform parametric analysis DSO. The command-line interface supported by this batch program is consistent with the command line used for current DSO jobs. desktopjob -help lists all available command-line options as shown below:

Command Line Syntax:

desktopjob.exe <options> <project-path-on-shared-drive>

Note that the project path can be to an archive file.

Options:

• -help

  Prints the help text.

• -cmd

  Specifies the command to run. Available choices: dso

• -ng

  Runs the analysis in non-graphical mode.

• -monitor

  Outputs progress and messages to standard output/error.

• -waitforlicense

  Queues the job until licenses are available.

• -preserve

  Preserves the local storage space of the distributed job for investigation into the job's run. If local storage directory (for example, the temp directory) is provisioned by scheduler, ensure it is also configured to preserve the job's local storage. This storage should be deleted manually.

• -batchoptions

  Overrides the Tools/Option entries through either a batchoptions file or batchoptions string.

  -batchoptions specific for Large Scale DSO include

  • LargeScaleDSO/FailedVarRetryCount: Retries failed variations.
  • FailedVarRetryCount: Instructs each task to re-simulate its failed variations when its assigned variations are finished. Input is a positive integer. Default is 0; Value-Zero(0) will disable the re-simulation for failed variations. Value ranges from 0 or positive integer.​
  • LargeScaleDSO/VarRedistribution: Redistribution batch option; ​Value 0 disables redistribution (default), and​ Value 1 enables redistribution.​ ​
  • LargeScaleDSO/RedistributionLimit: Specifies the minimum estimated remaining time (in minutes) for variations to redistribute to another task.​ Value must be a positive integer; the default is 3 minutes.

  Example:

  -batchoptions <config-file-on-shared-drive>

  -batchoptions "'name1'='val1' 'n2'='v2' "

• -jobid

  Specify a custom job ID for the job. The job's output is organized into a folder with job ID name. This parameter is ignored when run under a scheduler.

• -machinelist

In the context of Ansoft RSM:

Specify machines for distributed analysis. Machine list is specified either inline (as a comma separated machine names) or through a file. Multiple cores are specified by repeating the name of machine or by embedding number of cores in the machine name, using a colon separator.

Example 1:

-machinelist "list=m1,m1,m1,m2,m2,m3"

Example 2:

-machinelist "list=m1:3,m2:2,m3"

Example 3:

-machinelist "list=m1:1:3,m2:2:2"

Example 4:

-machinelist "file=machines.txt"

In the context of a scheduler such as LSF:

Specify the portion of total machines for distributed analysis. Use remaining for overhead or shared memory multiprocessing.

Manual Example:

-machinelist "Num=10"

Auto Example:

-machinelist "NumCores=40"

Note: In a scheduler environment, the machinelist argument can be omitted, but the results may not be as expected or desired. If there is no machinelist argument, the scheduler-assigned resources will be used, assuming one core per task. E.g., if the scheduler assigns 40 cores, there will be 40 variations solved in parallel, with only one core used per variation. Since one core per task may not make sense in many cases or may not be what the user desires, a warning message that machinelist is missing will be output, but the job will proceed.

• -auto

  Run the leaf jobs in auto mode. See Submitting Large Scale DSO Job Examples.

• -numdistributedvariations

  Specify the number of parallel leaf jobs to run in auto mode. Required when running under a scheduler. If not specified when not running under a scheduler, then the number of tasks for each machine must be specified in the machine list. The total number of tasks will be the number of parallel variations (leaf jobs) to run.

  Example:

  -machinelist NumCores=40 -auto -NumDistributedVariations 10

• -usefolderasinput

  Choose this option if the job's input represents an entire folder rather than just the project file.

• -maxfolderInMB

  Specify the maximum size input folder that is allowed for a valid job (in MBytes). By default, the maximum size allowed for input is 10MB. Specify a value of 0 to remove this size restriction and enable inputs of any size. This option applies when -usefolderasinput is used.

• -workdir

  Specifies the shared drive folder for status and result files generated by analysis. By default, the results folder of input project is used as the work directory.

• -mergecsv: [acrossDPs | singleDP | both]

  across DPs: Merge report csv files for all design-points (variations). One file is created per trace, across all variations.
  singleDP: Merge csv files within a single design-point (variation). One file is created per variation, per a set of traces that can be merged.
  both: Merge all traces that have the same primary sweep for all design-points (variations) into one csv file.
  Interpolation note: If primary sweep values are not uniformly spaced, mergecsv is enabled with traced values and are re-sampled uniformly using '-batchoptions' syntax as shown below:
  
  -batchoptions "'LargeScaleDSO/NumTracePoints'='PrimarySweepName:200'""
  -batchoptions "'LargeScaleDSO/NumTracePoints'='ReportName1:Trace1:100;ReportName1:Trace1:100;ReportName1:TraceName2:200'"

• -abort

  Abort a running job identified through the job's working directory. Example: -abort <projectresultsfolder-path>/<jobid>. For a complete discussion of methods for aborting jobs or specific tasks, see the discussion of Aborting a Large Scale DSO Simulation under Large Scale DSO for Parametric Analysis.

• -repackageresults

  Choose this option to add simulations results to the input archive file. Note that this option only applies if an archive file is provided as input.

• -generatearchive

After the simulation is complete, create an archive containing the project and results files. Note that this option only applies if project file is provided as input.

• -batchsolve

  Solves the specified parametric setup.

  Syntax for the setup:

  <design-name>:Optimetrics:<parametric-setup>

Below is an example of a desktop command to run a Large Scale DSO computation: