Calculations in Milliman Mind are parallelized using multiple threads of execution. The strategy used to spread the calculations into threads can be modified in the project settings under the category Run.

Different strategy can result in different run time. There is no definite way to determine the most efficient strategy without running the model, the reason being that chains of calculations are specific to each model.

There are 7 multicore strategies:

• Automatic (by default)
  • If the model is stochastic, the strategy used is Stochastic Optimized.
  • Otherwise if the model has instances, Mind automatically chooses the strategy Instances or Loops.
  • Otherwise, the system uses the strategy LoopOptimized for cells with loops and Classic for other cells.
• Classic
  This original strategy allocates elements into threads based on cell coordinates, loop and instance indexes.
• Instances or Loops
  If a cell has several instances, each instance is parallelized. If the cell has one or several loops, each index of loop is allocated to one thread. Otherwise, calculations of cells without dimension are launched on the main thread.
• Stochastic Optimized
  If a cell has stochastic simulations, all simulations of the cells are parallelized. Otherwise, calculations of cells without simulations are launched on the main thread.
• Loop Optimized
  If a cell has loop dimensions, each index of loop is parallelized. Otherwise, calculations of cells without loop dimensions are launched on the main thread.
• Instance Optimized
  If a cell has instances, each instance is parallelized. Otherwise, calculations of cells without instances are launched on the main thread.
• Cell Optimized
  Calculations of cells are allocated equally to threads based on their coordinates.