Parallel Meshing

You can use multiple cores to speed up volume mesh generation and create larger meshes than it is possible to create with a single core.

This feature is particularly useful when meshing large parts. Parallel meshes are conformal and of comparable quality to serial meshes. To use parallel meshing in a simulation, start the simulation in parallel mode.

The following meshers support parallel meshing:

• Polyhedral Mesher (including the underlying tetrahedral mesher)
• Tetrahedral Mesher
• Prism Layer Mesher (including the prism layer subsurface generation for the trimmed cell mesher. For the tetrahedral/polyhedral mesher, the subsurface generation is carried out in serial.)
• Trimmed Cell Mesher
• Advancing Layer Mesher (the prisms are first generated in serial, then the subsequent mesh pipeline executes in parallel.)
• Thin Mesher (the thin parts pre-processing runs in serial; the core volume mesh of the thin and bulk parts is generated in parallel.)

The following meshers do not support parallel meshing:

• Extruder Mesher
• Generalized Cylinder Mesher

For an average mesh (ten million cells or less), it is best to use between four and eight cores with one to two million cells per core. Using a large number of cores for a relatively small mesh reduces scalability and can result in slower performance. For polyhedral and tetrahedral meshes, a minimum of 250 thousand cells are needed per core. However, for trimmed cell meshes, you can use a minimum of 100 thousand cells per core. If the cell count for each core in parallel falls below the minimum number of cells per core, the trimmed cell mesher automatically uses fewer cores than are available and re-distributes the workload between these cores. This behavior ensures that there are a suitable number of cells for each core. When the total number of cells that are meshed in parallel falls below the minimum required for each core, the trimmed cell mesher automatically falls back to serial.

For distributed memory machines, the mesh size is limited by the number of cores and the amount of memory that is available to each core. Ensure that you have enough cores—and memory available to each core—for your entire mesh. The following table shows general memory requirement guidelines when meshing in parallel using the different mesh types. Memory requirement for some cases can fall outside these values.