Polyhedral modelling with exact arithmetic
Steven Fortune · 1995
We show that a boundary-based polyhedral solid modeller can be implemented using exact, multiprecision integer arithmetic with minimal performance overhead compared to a floating-point arithmetic implementation, Exact arithmetic guarantees numerical reliability with no additional algorithmic requirements.Exact arithmetic also simplifies some implementation details; for example, symbolic perturbation obviates degenerate-case handling.The performance cost of multiprecision arithmetic is reduced by appropriate tuningfor example, a floating-point filter eliminates many invocations of multiprecision arithmetic-and by algorithm design to minimize the arithmetic bit-length required by geometric primitives.