Parallel Part Meshing

You can control three mechanisms in Ansys Workbench that operate in a parallelized manner:

  • Remote Solve Manager Design Point updates. Refer to RSM Configuration.

  • Parallel Part meshing: Tools> Options> Number of CPUs for Parallel Part Meshing

  • Individual mesh methods (MultiZone Quad/Tri, Patch Independent Tetra, and MultiZone only): Tools Options > Number of CPUs for Meshing Methods

For the most efficient use of machine resources, it is important that the running processes do not over-saturate the processing cores or the available memory. You must allocate processing cores to each of these mechanisms in a way that provides the most benefit for your workflow. When parallel part meshing is invoked with the default number of CPUs, it automatically uses the cores of all available CPUs with a minimum of 2 gigabytes per CPU core.

Parallel Part Meshing Best Practices

Best practices include:

  • Know how many physical processing cores are available.

  • If you are using Remote Solve Manager (RSM), meshing is done serially. This option cannot be overridden.

  • For non-RSM Design Point updates, meshing is done serially by default. You can override this option by setting the option Number of CPUs for Parallel Part Meshing explicitly under Meshing > Advanced Options.

  • If you are using more than one processor for individual meshing methods, be sure to set a balance between the two meshing options if you are using methods that support the Number of CPUs for Meshing Methods option (MultiZone Quad/Tri, Patch Independent Tetra, and MultiZone).  They should be set to an explicit value greater than 0.

    For example, if you are using an 8 core system, setting Number of CPUs for Parallel Part Meshing to 3 and Number of CPUs for Mesh Methods to 3 will provide a good balance. If the mesh methods that you typically use support the Number of CPUs for Meshing Methods option, setting the Number of CPUs for Parallel Part Meshing to 2 and the Number of CPUs for Meshing Methods to 4 or 5 may potentially provide more benefit.

  • If you are using parallel part meshing only, you can set the Number of CPUs for Parallel Part Meshing to 0.  In such cases, the software uses as many cores as possible.

  • For Parallel Part meshing, you should turn off hyper-threading as this may lead to degradation of parallel performance.

  • Note that Parallel Part meshing does not support the following mesh controls:

    • Contact Sizing

    • Fracture

    • Mesh Match via Symmetry

    • Morph Service/Morphing

    • Pinch

    • Post Connection

    • Post-Inflation

    • Preview Surface Mesh/Preview Inflation

    • Retry

    • Refinement