In this paper, techniques for optimization of net algorithms describing parallel asynchronous computations and derived from cycling and branching behavioral descriptions are presented. The parallelization level of the algorithms is defined by a set of parallel operator pairs. The optimization techniques cover the two key steps of parallelization flow: the generation of an optimal initial set of parallel operator pairs to meet the constraints on the execution time or implementation cost, and the generation of the final set of pairs to solve the net algorithm existence problem. The quality of the proposed techniques is evaluated by experimental results. The techniques based on the minimization of the net algorithm critical paths estimated using the maximal weight cliques of the sequential and parallel operator graphs constitute the most efficient approach to the generation of the initial and final sets of parallel operator pairs.