Implementation implications on the performance of ANCF simulations

Implementation choices when writing multibody-dynamics software can be critically important to the overall performance. This is no different when applied to the Absolute Nodal Coordinate Formulation to include non-linear flexibility in the simulation. While reducing the number of floating point operations is important when developing efficient implementations, the architecture and constraints of the target hardware also need to be carefully considered. A modified calculation method of the internal forces with a new derivation and a new calculation method for the Jacobian of the internal forces is presented with both mathematical and software implementation details that attempt to include these hardware aspects into the implementation. Several different software implementations are compared running on physical hardware to demonstrate the cost of not optimizing the implementation and to show significant performance gains are possible when better aligning the implementation to the hardware architecture even while maintaining the same number of floating point operations.