This paper presents an implementation of motion optimization for robots. Functional optimization is converted to parameter optimization using B-splines. Many non-linear constraints are considered such as contacts stability, motors limitations and regularized collision avoidance. Semi-infinite constraints are simply solved by a fine enought discretization on a grid. Energy consumption is minimized. Motion optimization is solved with the IPOPT optimization package. For fast and reliable optimization convergence, gradients are computed. Joint friction is taken into account and regularized to obtain a smooth problem. Efficiency of the proposed software is demonstrated for three dynamic motions of the 30 degrees-of-freedom humanoid robot HRP-2: a kicking motion, a throwing motion and a lifting motion.