Conformational searching is a core task in inverse molecular kinematics. Algorithmic improvements affecting either the speed or quality of conformational searching will have a profound impact on applications including ligand-receptor docking, ab initio prediction of protein structure, and protein folding. In this talk, we focus on a specific geometry-constrained conformational searching problem, where some feature atoms have pre-specified target positions. Using Bezier subdivision, we present a method to locate and approximate the solutions of the equations derived from constraints on the feature atoms. The conformations corresponding to these solutions are all the conformations satisfying the target constraints. Three implementations of the subdivision method taking advantage of the sparsity of the coefficients of the polynomial equations are presented and the results are compared.