Step the world. This uses an iterative method that takes time on the order of m*N and memory on the order of m, where m is the total number of constraint rows and N is the number of iterations. For large systems this is a lot faster than dWorldStep, but it is less accurate.
QuickStep is great for stacks of objects especially when the auto-disable feature is used as well. However, it has poor accuracy for near-singular systems. Near-singular systems can occur when using high-friction contacts, motors, or certain articulated structures. For example, a robot with multiple legs sitting on the ground may be near-singular. There are ways to help overcome QuickStep's inaccuracy problems: - Increase CFM. - Reduce the number of contacts in your system (e.g. use the minimum number of contacts for the feet of a robot or creature). - Don't use excessive friction in the contacts. - Use contact slip if appropriate - Avoid kinematic loops (however, kinematic loops are inevitable in legged creatures). - Don't use excessive motor strength. - Use force-based motors instead of velocity-based motors. Increasing the number of QuickStep iterations may help a little bit, but it is not going to help much if your system is really near singular.